diff options
| author | Gabe Black <gblack@eecs.umich.edu> | 2010-10-31 00:07:20 -0700 |
|---|---|---|
| committer | Gabe Black <gblack@eecs.umich.edu> | 2010-10-31 00:07:20 -0700 |
| commit | 6f4bd2c1da0dc7783da87c4877a41332901199b2 (patch) | |
| tree | 99f2898e2b659338fd0b01d86eb9a4f8d981e21a /src/cpu | |
| parent | 373154a25afb1bed946e5a2a7cfd411e4bd7fad6 (diff) | |
| download | gem5-6f4bd2c1da0dc7783da87c4877a41332901199b2.tar.xz | |
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
Diffstat (limited to 'src/cpu')
68 files changed, 937 insertions, 1622 deletions
diff --git a/src/cpu/base_dyn_inst.hh b/src/cpu/base_dyn_inst.hh index e9b7daa4a..bc6f59407 100644 --- a/src/cpu/base_dyn_inst.hh +++ b/src/cpu/base_dyn_inst.hh @@ -38,6 +38,7 @@ #include <string> #include "arch/faults.hh" +#include "arch/utility.hh" #include "base/fast_alloc.hh" #include "base/trace.hh" #include "config/full_system.hh" @@ -241,36 +242,15 @@ class BaseDynInst : public FastAlloc, public RefCounted /** Records changes to result? */ bool recordResult; - /** PC of this instruction. */ - Addr PC; - - /** Micro PC of this instruction. */ - Addr microPC; - /** Did this instruction execute, or is it predicated false */ bool predicate; protected: - /** Next non-speculative PC. It is not filled in at fetch, but rather - * once the target of the branch is truly known (either decode or - * execute). - */ - Addr nextPC; - - /** Next non-speculative NPC. Target PC for Mips or Sparc. */ - Addr nextNPC; + /** PC state for this instruction. */ + TheISA::PCState pc; - /** Next non-speculative micro PC. */ - Addr nextMicroPC; - - /** Predicted next PC. */ - Addr predPC; - - /** Predicted next NPC. */ - Addr predNPC; - - /** Predicted next microPC */ - Addr predMicroPC; + /** Predicted PC state after this instruction. */ + TheISA::PCState predPC; /** If this is a branch that was predicted taken */ bool predTaken; @@ -386,27 +366,23 @@ class BaseDynInst : public FastAlloc, public RefCounted } /** BaseDynInst constructor given a binary instruction. * @param staticInst A StaticInstPtr to the underlying instruction. - * @param PC The PC of the instruction. - * @param pred_PC The predicted next PC. - * @param pred_NPC The predicted next NPC. + * @param pc The PC state for the instruction. + * @param predPC The predicted next PC state for the instruction. * @param seq_num The sequence number of the instruction. * @param cpu Pointer to the instruction's CPU. */ - BaseDynInst(StaticInstPtr staticInst, Addr PC, Addr NPC, Addr microPC, - Addr pred_PC, Addr pred_NPC, Addr pred_MicroPC, - InstSeqNum seq_num, ImplCPU *cpu); + BaseDynInst(StaticInstPtr staticInst, TheISA::PCState pc, + TheISA::PCState predPC, InstSeqNum seq_num, ImplCPU *cpu); /** BaseDynInst constructor given a binary instruction. * @param inst The binary instruction. - * @param PC The PC of the instruction. - * @param pred_PC The predicted next PC. - * @param pred_NPC The predicted next NPC. + * @param _pc The PC state for the instruction. + * @param _predPC The predicted next PC state for the instruction. * @param seq_num The sequence number of the instruction. * @param cpu Pointer to the instruction's CPU. */ - BaseDynInst(TheISA::ExtMachInst inst, Addr PC, Addr NPC, Addr microPC, - Addr pred_PC, Addr pred_NPC, Addr pred_MicroPC, - InstSeqNum seq_num, ImplCPU *cpu); + BaseDynInst(TheISA::ExtMachInst inst, TheISA::PCState pc, + TheISA::PCState predPC, InstSeqNum seq_num, ImplCPU *cpu); /** BaseDynInst constructor given a StaticInst pointer. * @param _staticInst The StaticInst for this BaseDynInst. @@ -443,45 +419,22 @@ class BaseDynInst : public FastAlloc, public RefCounted */ bool doneTargCalc() { return false; } - /** Returns the next PC. This could be the speculative next PC if it is - * called prior to the actual branch target being calculated. - */ - Addr readNextPC() { return nextPC; } - - /** Returns the next NPC. This could be the speculative next NPC if it is - * called prior to the actual branch target being calculated. - */ - Addr readNextNPC() - { -#if ISA_HAS_DELAY_SLOT - return nextNPC; -#else - return nextPC + sizeof(TheISA::MachInst); -#endif - } - - Addr readNextMicroPC() - { - return nextMicroPC; - } - /** Set the predicted target of this current instruction. */ - void setPredTarg(Addr predicted_PC, Addr predicted_NPC, - Addr predicted_MicroPC) + void setPredTarg(const TheISA::PCState &_predPC) { - predPC = predicted_PC; - predNPC = predicted_NPC; - predMicroPC = predicted_MicroPC; + predPC = _predPC; } + const TheISA::PCState &readPredTarg() { return predPC; } + /** Returns the predicted PC immediately after the branch. */ - Addr readPredPC() { return predPC; } + Addr predInstAddr() { return predPC.instAddr(); } /** Returns the predicted PC two instructions after the branch */ - Addr readPredNPC() { return predNPC; } + Addr predNextInstAddr() { return predPC.nextInstAddr(); } /** Returns the predicted micro PC after the branch */ - Addr readPredMicroPC() { return predMicroPC; } + Addr predMicroPC() { return predPC.microPC(); } /** Returns whether the instruction was predicted taken or not. */ bool readPredTaken() @@ -497,9 +450,9 @@ class BaseDynInst : public FastAlloc, public RefCounted /** Returns whether the instruction mispredicted. */ bool mispredicted() { - return readPredPC() != readNextPC() || - readPredNPC() != readNextNPC() || - readPredMicroPC() != readNextMicroPC(); + TheISA::PCState tempPC = pc; + TheISA::advancePC(tempPC, staticInst); + return !(tempPC == predPC); } // @@ -576,7 +529,8 @@ class BaseDynInst : public FastAlloc, public RefCounted OpClass opClass() const { return staticInst->opClass(); } /** Returns the branch target address. */ - Addr branchTarget() const { return staticInst->branchTarget(PC); } + TheISA::PCState branchTarget() const + { return staticInst->branchTarget(pc); } /** Returns the number of source registers. */ int8_t numSrcRegs() const { return staticInst->numSrcRegs(); } @@ -773,30 +727,20 @@ class BaseDynInst : public FastAlloc, public RefCounted /** Returns whether or not this instruction is squashed in the ROB. */ bool isSquashedInROB() const { return status[SquashedInROB]; } - /** Read the PC of this instruction. */ - const Addr readPC() const { return PC; } + /** Read the PC state of this instruction. */ + const TheISA::PCState pcState() const { return pc; } - /**Read the micro PC of this instruction. */ - const Addr readMicroPC() const { return microPC; } + /** Set the PC state of this instruction. */ + const void pcState(const TheISA::PCState &val) { pc = val; } - /** Set the next PC of this instruction (its actual target). */ - void setNextPC(Addr val) - { - nextPC = val; - } + /** Read the PC of this instruction. */ + const Addr instAddr() const { return pc.instAddr(); } - /** Set the next NPC of this instruction (the target in Mips or Sparc).*/ - void setNextNPC(Addr val) - { -#if ISA_HAS_DELAY_SLOT - nextNPC = val; -#endif - } + /** Read the PC of the next instruction. */ + const Addr nextInstAddr() const { return pc.nextInstAddr(); } - void setNextMicroPC(Addr val) - { - nextMicroPC = val; - } + /**Read the micro PC of this instruction. */ + const Addr microPC() const { return pc.microPC(); } bool readPredicate() { @@ -895,7 +839,7 @@ BaseDynInst<Impl>::readBytes(Addr addr, uint8_t *data, unsigned size, unsigned flags) { reqMade = true; - Request *req = new Request(asid, addr, size, flags, this->PC, + Request *req = new Request(asid, addr, size, flags, this->pc.instAddr(), thread->contextId(), threadNumber); Request *sreqLow = NULL; @@ -956,7 +900,7 @@ BaseDynInst<Impl>::writeBytes(uint8_t *data, unsigned size, } reqMade = true; - Request *req = new Request(asid, addr, size, flags, this->PC, + Request *req = new Request(asid, addr, size, flags, this->pc.instAddr(), thread->contextId(), threadNumber); Request *sreqLow = NULL; diff --git a/src/cpu/base_dyn_inst_impl.hh b/src/cpu/base_dyn_inst_impl.hh index 4fb8490c1..923b204ce 100644 --- a/src/cpu/base_dyn_inst_impl.hh +++ b/src/cpu/base_dyn_inst_impl.hh @@ -62,32 +62,14 @@ my_hash_t thishash; template <class Impl> BaseDynInst<Impl>::BaseDynInst(StaticInstPtr _staticInst, - Addr inst_PC, Addr inst_NPC, - Addr inst_MicroPC, - Addr pred_PC, Addr pred_NPC, - Addr pred_MicroPC, + TheISA::PCState _pc, TheISA::PCState _predPC, InstSeqNum seq_num, ImplCPU *cpu) : staticInst(_staticInst), traceData(NULL), cpu(cpu) { seqNum = seq_num; - bool nextIsMicro = - staticInst->isMicroop() && !staticInst->isLastMicroop(); - - PC = inst_PC; - microPC = inst_MicroPC; - if (nextIsMicro) { - nextPC = inst_PC; - nextNPC = inst_NPC; - nextMicroPC = microPC + 1; - } else { - nextPC = inst_NPC; - nextNPC = nextPC + sizeof(TheISA::MachInst); - nextMicroPC = 0; - } - predPC = pred_PC; - predNPC = pred_NPC; - predMicroPC = pred_MicroPC; + pc = _pc; + predPC = _predPC; predTaken = false; initVars(); @@ -95,32 +77,14 @@ BaseDynInst<Impl>::BaseDynInst(StaticInstPtr _staticInst, template <class Impl> BaseDynInst<Impl>::BaseDynInst(TheISA::ExtMachInst inst, - Addr inst_PC, Addr inst_NPC, - Addr inst_MicroPC, - Addr pred_PC, Addr pred_NPC, - Addr pred_MicroPC, + TheISA::PCState _pc, TheISA::PCState _predPC, InstSeqNum seq_num, ImplCPU *cpu) - : staticInst(inst, inst_PC), traceData(NULL), cpu(cpu) + : staticInst(inst, _pc.instAddr()), traceData(NULL), cpu(cpu) { seqNum = seq_num; - bool nextIsMicro = - staticInst->isMicroop() && !staticInst->isLastMicroop(); - - PC = inst_PC; - microPC = inst_MicroPC; - if (nextIsMicro) { - nextPC = inst_PC; - nextNPC = inst_NPC; - nextMicroPC = microPC + 1; - } else { - nextPC = inst_NPC; - nextNPC = nextPC + sizeof(TheISA::MachInst); - nextMicroPC = 0; - } - predPC = pred_PC; - predNPC = pred_NPC; - predMicroPC = pred_MicroPC; + pc = _pc; + predPC = _predPC; predTaken = false; initVars(); @@ -301,8 +265,8 @@ template <class Impl> void BaseDynInst<Impl>::dump() { - cprintf("T%d : %#08d `", threadNumber, PC); - std::cout << staticInst->disassemble(PC); + cprintf("T%d : %#08d `", threadNumber, pc.instAddr()); + std::cout << staticInst->disassemble(pc.instAddr()); cprintf("'\n"); } @@ -311,8 +275,8 @@ void BaseDynInst<Impl>::dump(std::string &outstring) { std::ostringstream s; - s << "T" << threadNumber << " : 0x" << PC << " " - << staticInst->disassemble(PC); + s << "T" << threadNumber << " : 0x" << pc.instAddr() << " " + << staticInst->disassemble(pc.instAddr()); outstring = s.str(); } diff --git a/src/cpu/checker/cpu.hh b/src/cpu/checker/cpu.hh index a2f44106d..df5d8209b 100644 --- a/src/cpu/checker/cpu.hh +++ b/src/cpu/checker/cpu.hh @@ -241,13 +241,9 @@ class CheckerCPU : public BaseCPU result.integer = val; } - uint64_t readPC() { return thread->readPC(); } + uint64_t instAddr() { return thread->instAddr(); } - uint64_t readNextPC() { return thread->readNextPC(); } - - void setNextPC(uint64_t val) { - thread->setNextPC(val); - } + uint64_t nextInstAddr() { return thread->nextInstAddr(); } MiscReg readMiscRegNoEffect(int misc_reg) { diff --git a/src/cpu/exetrace.cc b/src/cpu/exetrace.cc index 4bab778ba..f8b25ef73 100644 --- a/src/cpu/exetrace.cc +++ b/src/cpu/exetrace.cc @@ -72,10 +72,7 @@ Trace::ExeTracerRecord::traceInst(StaticInstPtr inst, bool ran) std::string sym_str; Addr sym_addr; - Addr cur_pc = PC; -#if THE_ISA == ARM_ISA - cur_pc &= ~PcModeMask; -#endif + Addr cur_pc = pc.instAddr(); if (debugSymbolTable && IsOn(ExecSymbol) #if FULL_SYSTEM @@ -85,13 +82,12 @@ Trace::ExeTracerRecord::traceInst(StaticInstPtr inst, bool ran) if (cur_pc != sym_addr) sym_str += csprintf("+%d",cur_pc - sym_addr); outs << "@" << sym_str; - } - else { + } else { outs << "0x" << hex << cur_pc; } if (inst->isMicroop()) { - outs << "." << setw(2) << dec << upc; + outs << "." << setw(2) << dec << pc.microPC(); } else { outs << " "; } diff --git a/src/cpu/exetrace.hh b/src/cpu/exetrace.hh index 1982595eb..65950728b 100644 --- a/src/cpu/exetrace.hh +++ b/src/cpu/exetrace.hh @@ -47,10 +47,10 @@ class ExeTracerRecord : public InstRecord { public: ExeTracerRecord(Tick _when, ThreadContext *_thread, - const StaticInstPtr _staticInst, Addr _pc, bool spec, - const StaticInstPtr _macroStaticInst = NULL, MicroPC _upc = 0) + const StaticInstPtr _staticInst, TheISA::PCState _pc, + bool spec, const StaticInstPtr _macroStaticInst = NULL) : InstRecord(_when, _thread, _staticInst, _pc, spec, - _macroStaticInst, _upc) + _macroStaticInst) { } @@ -69,8 +69,8 @@ class ExeTracer : public InstTracer InstRecord * getInstRecord(Tick when, ThreadContext *tc, - const StaticInstPtr staticInst, Addr pc, - const StaticInstPtr macroStaticInst = NULL, MicroPC upc = 0) + const StaticInstPtr staticInst, TheISA::PCState pc, + const StaticInstPtr macroStaticInst = NULL) { if (!IsOn(ExecEnable)) return NULL; @@ -82,7 +82,7 @@ class ExeTracer : public InstTracer return NULL; return new ExeTracerRecord(when, tc, - staticInst, pc, tc->misspeculating(), macroStaticInst, upc); + staticInst, pc, tc->misspeculating(), macroStaticInst); } }; diff --git a/src/cpu/inorder/comm.hh b/src/cpu/inorder/comm.hh index f1b3cacac..386046d1c 100644 --- a/src/cpu/inorder/comm.hh +++ b/src/cpu/inorder/comm.hh @@ -75,8 +75,8 @@ struct TimeStruct { // struct as it is used pretty frequently. bool branchMispredict; bool branchTaken; - uint64_t mispredPC; - uint64_t nextPC; + Addr mispredPC; + TheISA::PCState nextPC; unsigned branchCount; diff --git a/src/cpu/inorder/cpu.cc b/src/cpu/inorder/cpu.cc index 5d42ba559..6cd938dc6 100644 --- a/src/cpu/inorder/cpu.cc +++ b/src/cpu/inorder/cpu.cc @@ -988,47 +988,6 @@ InOrderCPU::getPipeStage(int stage_num) } uint64_t -InOrderCPU::readPC(ThreadID tid) -{ - return PC[tid]; -} - - -void -InOrderCPU::setPC(Addr new_PC, ThreadID tid) -{ - PC[tid] = new_PC; -} - - -uint64_t -InOrderCPU::readNextPC(ThreadID tid) -{ - return nextPC[tid]; -} - - -void -InOrderCPU::setNextPC(uint64_t new_NPC, ThreadID tid) -{ - nextPC[tid] = new_NPC; -} - - -uint64_t -InOrderCPU::readNextNPC(ThreadID tid) -{ - return nextNPC[tid]; -} - - -void -InOrderCPU::setNextNPC(uint64_t new_NNPC, ThreadID tid) -{ - nextNPC[tid] = new_NNPC; -} - -uint64_t InOrderCPU::readIntReg(int reg_idx, ThreadID tid) { return intRegs[tid][reg_idx]; @@ -1156,15 +1115,12 @@ InOrderCPU::instDone(DynInstPtr inst, ThreadID tid) // Set the CPU's PCs - This contributes to the precise state of the CPU // which can be used when restoring a thread to the CPU after after any // type of context switching activity (fork, exception, etc.) - setPC(inst->readPC(), tid); - setNextPC(inst->readNextPC(), tid); - setNextNPC(inst->readNextNPC(), tid); + pcState(inst->pcState(), tid); if (inst->isControl()) { thread[tid]->lastGradIsBranch = true; - thread[tid]->lastBranchPC = inst->readPC(); - thread[tid]->lastBranchNextPC = inst->readNextPC(); - thread[tid]->lastBranchNextNPC = inst->readNextNPC(); + thread[tid]->lastBranchPC = inst->pcState(); + TheISA::advancePC(thread[tid]->lastBranchPC, inst->staticInst); } else { thread[tid]->lastGradIsBranch = false; } @@ -1236,15 +1192,15 @@ InOrderCPU::addToRemoveList(DynInstPtr &inst) { removeInstsThisCycle = true; if (!inst->isRemoveList()) { - DPRINTF(InOrderCPU, "Pushing instruction [tid:%i] PC %#x " + DPRINTF(InOrderCPU, "Pushing instruction [tid:%i] PC %s " "[sn:%lli] to remove list\n", - inst->threadNumber, inst->readPC(), inst->seqNum); + inst->threadNumber, inst->pcState(), inst->seqNum); inst->setRemoveList(); removeList.push(inst->getInstListIt()); } else { - DPRINTF(InOrderCPU, "Ignoring instruction removal for [tid:%i] PC %#x " + DPRINTF(InOrderCPU, "Ignoring instruction removal for [tid:%i] PC %s " "[sn:%lli], already remove list\n", - inst->threadNumber, inst->readPC(), inst->seqNum); + inst->threadNumber, inst->pcState(), inst->seqNum); } } @@ -1252,23 +1208,23 @@ InOrderCPU::addToRemoveList(DynInstPtr &inst) void InOrderCPU::removeInst(DynInstPtr &inst) { - DPRINTF(InOrderCPU, "Removing graduated instruction [tid:%i] PC %#x " + DPRINTF(InOrderCPU, "Removing graduated instruction [tid:%i] PC %s " "[sn:%lli]\n", - inst->threadNumber, inst->readPC(), inst->seqNum); + inst->threadNumber, inst->pcState(), inst->seqNum); removeInstsThisCycle = true; // Remove the instruction. if (!inst->isRemoveList()) { - DPRINTF(InOrderCPU, "Pushing instruction [tid:%i] PC %#x " + DPRINTF(InOrderCPU, "Pushing instruction [tid:%i] PC %s " "[sn:%lli] to remove list\n", - inst->threadNumber, inst->readPC(), inst->seqNum); + inst->threadNumber, inst->pcState(), inst->seqNum); inst->setRemoveList(); removeList.push(inst->getInstListIt()); } else { - DPRINTF(InOrderCPU, "Ignoring instruction removal for [tid:%i] PC %#x " + DPRINTF(InOrderCPU, "Ignoring instruction removal for [tid:%i] PC %s " "[sn:%lli], already on remove list\n", - inst->threadNumber, inst->readPC(), inst->seqNum); + inst->threadNumber, inst->pcState(), inst->seqNum); } } @@ -1307,24 +1263,24 @@ InOrderCPU::squashInstIt(const ListIt &instIt, ThreadID tid) { if ((*instIt)->threadNumber == tid) { DPRINTF(InOrderCPU, "Squashing instruction, " - "[tid:%i] [sn:%lli] PC %#x\n", + "[tid:%i] [sn:%lli] PC %s\n", (*instIt)->threadNumber, (*instIt)->seqNum, - (*instIt)->readPC()); + (*instIt)->pcState()); (*instIt)->setSquashed(); if (!(*instIt)->isRemoveList()) { - DPRINTF(InOrderCPU, "Pushing instruction [tid:%i] PC %#x " + DPRINTF(InOrderCPU, "Pushing instruction [tid:%i] PC %s " "[sn:%lli] to remove list\n", - (*instIt)->threadNumber, (*instIt)->readPC(), + (*instIt)->threadNumber, (*instIt)->pcState(), (*instIt)->seqNum); (*instIt)->setRemoveList(); removeList.push(instIt); } else { DPRINTF(InOrderCPU, "Ignoring instruction removal for [tid:%i]" - " PC %#x [sn:%lli], already on remove list\n", - (*instIt)->threadNumber, (*instIt)->readPC(), + " PC %s [sn:%lli], already on remove list\n", + (*instIt)->threadNumber, (*instIt)->pcState(), (*instIt)->seqNum); } @@ -1338,10 +1294,10 @@ InOrderCPU::cleanUpRemovedInsts() { while (!removeList.empty()) { DPRINTF(InOrderCPU, "Removing instruction, " - "[tid:%i] [sn:%lli] PC %#x\n", + "[tid:%i] [sn:%lli] PC %s\n", (*removeList.front())->threadNumber, (*removeList.front())->seqNum, - (*removeList.front())->readPC()); + (*removeList.front())->pcState()); DynInstPtr inst = *removeList.front(); ThreadID tid = inst->threadNumber; @@ -1417,9 +1373,10 @@ InOrderCPU::dumpInsts() cprintf("Dumping Instruction List\n"); while (inst_list_it != instList[0].end()) { - cprintf("Instruction:%i\nPC:%#x\n[tid:%i]\n[sn:%lli]\nIssued:%i\n" + cprintf("Instruction:%i\nPC:%s\n[tid:%i]\n[sn:%lli]\nIssued:%i\n" "Squashed:%i\n\n", - num, (*inst_list_it)->readPC(), (*inst_list_it)->threadNumber, + num, (*inst_list_it)->pcState(), + (*inst_list_it)->threadNumber, (*inst_list_it)->seqNum, (*inst_list_it)->isIssued(), (*inst_list_it)->isSquashed()); inst_list_it++; diff --git a/src/cpu/inorder/cpu.hh b/src/cpu/inorder/cpu.hh index abe24d6ed..df6617d56 100644 --- a/src/cpu/inorder/cpu.hh +++ b/src/cpu/inorder/cpu.hh @@ -273,9 +273,7 @@ class InOrderCPU : public BaseCPU PipelineStage *pipelineStage[ThePipeline::NumStages]; /** Program Counters */ - TheISA::IntReg PC[ThePipeline::MaxThreads]; - TheISA::IntReg nextPC[ThePipeline::MaxThreads]; - TheISA::IntReg nextNPC[ThePipeline::MaxThreads]; + TheISA::PCState pc[ThePipeline::MaxThreads]; /** The Register File for the CPU */ union { @@ -471,22 +469,22 @@ class InOrderCPU : public BaseCPU ThreadID tid); /** Reads the commit PC of a specific thread. */ - uint64_t readPC(ThreadID tid); + TheISA::PCState + pcState(ThreadID tid) + { + return pc[tid]; + } /** Sets the commit PC of a specific thread. */ - void setPC(Addr new_PC, ThreadID tid); - - /** Reads the next PC of a specific thread. */ - uint64_t readNextPC(ThreadID tid); - - /** Sets the next PC of a specific thread. */ - void setNextPC(uint64_t val, ThreadID tid); - - /** Reads the next NPC of a specific thread. */ - uint64_t readNextNPC(ThreadID tid); + void + pcState(const TheISA::PCState &newPC, ThreadID tid) + { + pc[tid] = newPC; + } - /** Sets the next NPC of a specific thread. */ - void setNextNPC(uint64_t val, ThreadID tid); + Addr instAddr(ThreadID tid) { return pc[tid].instAddr(); } + Addr nextInstAddr(ThreadID tid) { return pc[tid].nextInstAddr(); } + MicroPC microPC(ThreadID tid) { return pc[tid].microPC(); } /** Function to add instruction onto the head of the list of the * instructions. Used when new instructions are fetched. diff --git a/src/cpu/inorder/first_stage.cc b/src/cpu/inorder/first_stage.cc index 658ce37d3..ae458c604 100644 --- a/src/cpu/inorder/first_stage.cc +++ b/src/cpu/inorder/first_stage.cc @@ -84,8 +84,8 @@ FirstStage::squash(InstSeqNum squash_seq_num, ThreadID tid) break; } DPRINTF(InOrderStage, "[tid:%i]: Removing instruction, [sn:%i] " - "PC %08p.\n", tid, insts[tid].front()->seqNum, - insts[tid].front()->PC); + "PC %s.\n", tid, insts[tid].front()->seqNum, + insts[tid].front()->pc); insts[tid].pop(); } diff --git a/src/cpu/inorder/inorder_dyn_inst.cc b/src/cpu/inorder/inorder_dyn_inst.cc index 2465744e5..f672082f3 100644 --- a/src/cpu/inorder/inorder_dyn_inst.cc +++ b/src/cpu/inorder/inorder_dyn_inst.cc @@ -47,17 +47,16 @@ using namespace std; using namespace TheISA; using namespace ThePipeline; -InOrderDynInst::InOrderDynInst(TheISA::ExtMachInst machInst, Addr inst_PC, - Addr pred_PC, InstSeqNum seq_num, - InOrderCPU *cpu) - : staticInst(machInst, inst_PC), traceData(NULL), cpu(cpu) +InOrderDynInst::InOrderDynInst(TheISA::ExtMachInst machInst, + const TheISA::PCState &instPC, + const TheISA::PCState &_predPC, + InstSeqNum seq_num, InOrderCPU *cpu) + : staticInst(machInst, instPC.instAddr()), traceData(NULL), cpu(cpu) { seqNum = seq_num; - PC = inst_PC; - nextPC = PC + sizeof(MachInst); - nextNPC = nextPC + sizeof(MachInst); - predPC = pred_PC; + pc = instPC; + predPC = _predPC; initVars(); } @@ -94,7 +93,7 @@ int InOrderDynInst::instcount = 0; void InOrderDynInst::setMachInst(ExtMachInst machInst) { - staticInst = StaticInst::decode(machInst, PC); + staticInst = StaticInst::decode(machInst, pc.instAddr()); for (int i = 0; i < this->staticInst->numDestRegs(); i++) { _destRegIdx[i] = this->staticInst->destRegIdx(i); @@ -747,8 +746,8 @@ InOrderDynInst::write(int32_t data, Addr addr, unsigned flags, uint64_t *res) void InOrderDynInst::dump() { - cprintf("T%d : %#08d `", threadNumber, PC); - cout << staticInst->disassemble(PC); + cprintf("T%d : %#08d `", threadNumber, pc.instAddr()); + cout << staticInst->disassemble(pc.instAddr()); cprintf("'\n"); } @@ -756,8 +755,8 @@ void InOrderDynInst::dump(std::string &outstring) { std::ostringstream s; - s << "T" << threadNumber << " : 0x" << PC << " " - << staticInst->disassemble(PC); + s << "T" << threadNumber << " : " << pc << " " + << staticInst->disassemble(pc.instAddr()); outstring = s.str(); } diff --git a/src/cpu/inorder/inorder_dyn_inst.hh b/src/cpu/inorder/inorder_dyn_inst.hh index 0d42f4696..105e37657 100644 --- a/src/cpu/inorder/inorder_dyn_inst.hh +++ b/src/cpu/inorder/inorder_dyn_inst.hh @@ -41,6 +41,7 @@ #include "arch/isa_traits.hh" #include "arch/mt.hh" #include "arch/types.hh" +#include "arch/utility.hh" #include "base/fast_alloc.hh" #include "base/trace.hh" #include "base/types.hh" @@ -108,12 +109,13 @@ class InOrderDynInst : public FastAlloc, public RefCounted /** BaseDynInst constructor given a binary instruction. * @param inst The binary instruction. * @param PC The PC of the instruction. - * @param pred_PC The predicted next PC. + * @param predPC The predicted next PC. * @param seq_num The sequence number of the instruction. * @param cpu Pointer to the instruction's CPU. */ - InOrderDynInst(ExtMachInst inst, Addr PC, Addr pred_PC, InstSeqNum seq_num, - InOrderCPU *cpu); + InOrderDynInst(ExtMachInst inst, const TheISA::PCState &PC, + const TheISA::PCState &predPC, InstSeqNum seq_num, + InOrderCPU *cpu); /** BaseDynInst constructor given a binary instruction. * @param seq_num The sequence number of the instruction. @@ -269,28 +271,10 @@ class InOrderDynInst : public FastAlloc, public RefCounted InstResult instResult[MaxInstDestRegs]; /** PC of this instruction. */ - Addr PC; - - /** Next non-speculative PC. It is not filled in at fetch, but rather - * once the target of the branch is truly known (either decode or - * execute). - */ - Addr nextPC; - - /** Next next non-speculative PC. It is not filled in at fetch, but rather - * once the target of the branch is truly known (either decode or - * execute). - */ - Addr nextNPC; + TheISA::PCState pc; /** Predicted next PC. */ - Addr predPC; - - /** Predicted next NPC. */ - Addr predNPC; - - /** Predicted next microPC */ - Addr predMicroPC; + TheISA::PCState predPC; /** Address to fetch from */ Addr fetchAddr; @@ -540,33 +524,14 @@ class InOrderDynInst : public FastAlloc, public RefCounted // //////////////////////////////////////////////////////////// /** Read the PC of this instruction. */ - const Addr readPC() const { return PC; } + const TheISA::PCState &pcState() const { return pc; } /** Sets the PC of this instruction. */ - void setPC(Addr pc) { PC = pc; } - - /** Returns the next PC. This could be the speculative next PC if it is - * called prior to the actual branch target being calculated. - */ - Addr readNextPC() { return nextPC; } + void pcState(const TheISA::PCState &_pc) { pc = _pc; } - /** Set the next PC of this instruction (its actual target). */ - void setNextPC(uint64_t val) { nextPC = val; } - - /** Returns the next NPC. This could be the speculative next NPC if it is - * called prior to the actual branch target being calculated. - */ - Addr readNextNPC() - { -#if ISA_HAS_DELAY_SLOT - return nextNPC; -#else - return nextPC + sizeof(TheISA::MachInst); -#endif - } - - /** Set the next PC of this instruction (its actual target). */ - void setNextNPC(uint64_t val) { nextNPC = val; } + const Addr instAddr() { return pc.instAddr(); } + const Addr nextInstAddr() { return pc.nextInstAddr(); } + const MicroPC microPC() { return pc.microPC(); } //////////////////////////////////////////////////////////// // @@ -574,38 +539,36 @@ class InOrderDynInst : public FastAlloc, public RefCounted // //////////////////////////////////////////////////////////// /** Set the predicted target of this current instruction. */ - void setPredTarg(Addr predicted_PC) { predPC = predicted_PC; } + void setPredTarg(const TheISA::PCState &predictedPC) + { predPC = predictedPC; } /** Returns the predicted target of the branch. */ - Addr readPredTarg() { return predPC; } + TheISA::PCState readPredTarg() { return predPC; } /** Returns the predicted PC immediately after the branch. */ - Addr readPredPC() { return predPC; } + Addr predInstAddr() { return predPC.instAddr(); } /** Returns the predicted PC two instructions after the branch */ - Addr readPredNPC() { return predNPC; } + Addr predNextInstAddr() { return predPC.nextInstAddr(); } /** Returns the predicted micro PC after the branch */ - Addr readPredMicroPC() { return predMicroPC; } + Addr readPredMicroPC() { return predPC.microPC(); } /** Returns whether the instruction was predicted taken or not. */ bool predTaken() { return predictTaken; } /** Returns whether the instruction mispredicted. */ - bool mispredicted() + bool + mispredicted() { -#if ISA_HAS_DELAY_SLOT - return predPC != nextNPC; -#else - return predPC != nextPC; -#endif + TheISA::PCState nextPC = pc; + TheISA::advancePC(nextPC, staticInst); + return !(nextPC == predPC); } - /** Returns whether the instruction mispredicted. */ - bool mistargeted() { return predPC != nextNPC; } - /** Returns the branch target address. */ - Addr branchTarget() const { return staticInst->branchTarget(PC); } + TheISA::PCState branchTarget() const + { return staticInst->branchTarget(pc); } /** Checks whether or not this instruction has had its branch target * calculated yet. For now it is not utilized and is hacked to be diff --git a/src/cpu/inorder/inorder_trace.cc b/src/cpu/inorder/inorder_trace.cc index 90c94a4f5..70a947671 100644 --- a/src/cpu/inorder/inorder_trace.cc +++ b/src/cpu/inorder/inorder_trace.cc @@ -70,15 +70,15 @@ InOrderTrace::getInstRecord(unsigned num_stages, bool stage_tracing, if (!Trace::enabled) return NULL; - return new InOrderTraceRecord(num_stages, stage_tracing, tc); + return new InOrderTraceRecord(num_stages, stage_tracing, tc, 0); } InOrderTraceRecord * InOrderTrace::getInstRecord(Tick when, ThreadContext *tc, - const StaticInstPtr staticInst, Addr pc, - const StaticInstPtr macroStaticInst, MicroPC upc) + const StaticInstPtr staticInst, TheISA::PCState _pc, + const StaticInstPtr macroStaticInst) { - return new InOrderTraceRecord(ThePipeline::NumStages, true, tc); + return new InOrderTraceRecord(ThePipeline::NumStages, true, tc, _pc); } /* namespace Trace */ } diff --git a/src/cpu/inorder/inorder_trace.hh b/src/cpu/inorder/inorder_trace.hh index ccc868f15..fd1427500 100644 --- a/src/cpu/inorder/inorder_trace.hh +++ b/src/cpu/inorder/inorder_trace.hh @@ -47,8 +47,8 @@ class InOrderTraceRecord : public ExeTracerRecord { public: InOrderTraceRecord(unsigned num_stages, bool _stage_tracing, - ThreadContext *_thread, bool spec = false) - : ExeTracerRecord(0, _thread, NULL, 0, spec) + ThreadContext *_thread, TheISA::PCState _pc, bool spec = false) + : ExeTracerRecord(0, _thread, NULL, _pc, spec) { stageTrace = _stage_tracing; stageCycle.resize(num_stages); @@ -75,7 +75,8 @@ class InOrderTraceRecord : public ExeTracerRecord { staticInst = _staticInst; } - void setPC(Addr _pc) { PC = _pc; } + + void setPC(TheISA::PCState _pc) { pc = _pc; } }; class InOrderTrace : public InstTracer @@ -87,9 +88,9 @@ class InOrderTrace : public InstTracer InOrderTraceRecord * getInstRecord(unsigned num_stages, bool stage_tracing, ThreadContext *tc); - virtual InOrderTraceRecord *getInstRecord(Tick when, ThreadContext *tc, - const StaticInstPtr staticInst, Addr pc, - const StaticInstPtr macroStaticInst = NULL, MicroPC upc = 0); + InOrderTraceRecord *getInstRecord(Tick when, ThreadContext *tc, + const StaticInstPtr staticInst, TheISA::PCState pc, + const StaticInstPtr macroStaticInst = NULL); }; /* namespace Trace */ } diff --git a/src/cpu/inorder/pipeline_stage.cc b/src/cpu/inorder/pipeline_stage.cc index f0b48d6f6..dc36965b0 100644 --- a/src/cpu/inorder/pipeline_stage.cc +++ b/src/cpu/inorder/pipeline_stage.cc @@ -350,27 +350,21 @@ PipelineStage::squashDueToBranch(DynInstPtr &inst, ThreadID tid) toPrevStages->stageInfo[stageNum][tid].squash = true; toPrevStages->stageInfo[stageNum][tid].nextPC = inst->readPredTarg(); + toPrevStages->stageInfo[stageNum][tid].branchTaken = + inst->pcState().branching(); #if ISA_HAS_DELAY_SLOT - toPrevStages->stageInfo[stageNum][tid].branchTaken = - inst->readNextNPC() != - (inst->readNextPC() + sizeof(TheISA::MachInst)); - - toPrevStages->stageInfo[stageNum][tid].bdelayDoneSeqNum = + toPrevStages->stageInfo[stageNum][tid].bdelayDoneSeqNum = inst->bdelaySeqNum; InstSeqNum squash_seq_num = inst->bdelaySeqNum; #else - toPrevStages->stageInfo[stageNum][tid].branchTaken = - inst->readNextPC() != - (inst->readPC() + sizeof(TheISA::MachInst)); - toPrevStages->stageInfo[stageNum][tid].bdelayDoneSeqNum = inst->seqNum; InstSeqNum squash_seq_num = inst->seqNum; #endif DPRINTF(InOrderStage, "Target being re-set to %08p\n", - inst->readPredTarg()); + inst->predInstAddr()); DPRINTF(InOrderStage, "[tid:%i]: Squashing after [sn:%i], " "due to [sn:%i] branch.\n", tid, squash_seq_num, inst->seqNum); @@ -398,10 +392,10 @@ PipelineStage::squashPrevStageInsts(InstSeqNum squash_seq_num, ThreadID tid) prevStage->insts[i]->seqNum > squash_seq_num) { // Change Comment to Annulling previous instruction DPRINTF(InOrderStage, "[tid:%i]: Squashing instruction, " - "[sn:%i] PC %08p.\n", + "[sn:%i] PC %s.\n", tid, prevStage->insts[i]->seqNum, - prevStage->insts[i]->readPC()); + prevStage->insts[i]->pcState()); prevStage->insts[i]->setSquashed(); prevStage->insts[i] = cpu->dummyBufferInst; @@ -429,8 +423,8 @@ PipelineStage::squash(InstSeqNum squash_seq_num, ThreadID tid) break; } DPRINTF(InOrderStage, "[tid:%i]: Removing instruction, [sn:%i] " - " PC %08p.\n", tid, skidBuffer[tid].front()->seqNum, - skidBuffer[tid].front()->PC); + " PC %s.\n", tid, skidBuffer[tid].front()->seqNum, + skidBuffer[tid].front()->pc); skidBuffer[tid].pop(); } @@ -488,8 +482,8 @@ PipelineStage::skidInsert(ThreadID tid) assert(tid == inst->threadNumber); - DPRINTF(InOrderStage,"[tid:%i]: Inserting [sn:%lli] PC:%#x into stage " - "skidBuffer %i\n", tid, inst->seqNum, inst->readPC(), + DPRINTF(InOrderStage,"[tid:%i]: Inserting [sn:%lli] PC:%s into stage " + "skidBuffer %i\n", tid, inst->seqNum, inst->pcState(), inst->threadNumber); skidBuffer[tid].push(inst); @@ -571,9 +565,9 @@ PipelineStage::activateThread(ThreadID tid) } else { DynInstPtr inst = switchedOutBuffer[tid]; - DPRINTF(InOrderStage,"[tid:%i]: Re-Inserting [sn:%lli] PC:%#x into" + DPRINTF(InOrderStage,"[tid:%i]: Re-Inserting [sn:%lli] PC:%s into" " stage skidBuffer %i\n", tid, inst->seqNum, - inst->readPC(), inst->threadNumber); + inst->pcState(), inst->threadNumber); // Make instruction available for pipeline processing skidBuffer[tid].push(inst); @@ -895,13 +889,12 @@ PipelineStage::processInsts(ThreadID tid) inst = insts_to_stage.front(); DPRINTF(InOrderStage, "[tid:%u]: Processing instruction [sn:%lli] " - "with PC %#x\n", - tid, inst->seqNum, inst->readPC()); + "with PC %s\n", tid, inst->seqNum, inst->pcState()); if (inst->isSquashed()) { - DPRINTF(InOrderStage, "[tid:%u]: Instruction %i with PC %#x is " + DPRINTF(InOrderStage, "[tid:%u]: Instruction %i with PC %s is " "squashed, skipping.\n", - tid, inst->seqNum, inst->readPC()); + tid, inst->seqNum, inst->pcState()); insts_to_stage.pop(); @@ -1001,8 +994,8 @@ PipelineStage::processInstSchedule(DynInstPtr inst,int &reqs_processed) switchedOutValid[tid] = true; // Remove Thread From Pipeline & Resource Pool - inst->squashingStage = stageNum; - inst->bdelaySeqNum = inst->seqNum; + inst->squashingStage = stageNum; + inst->bdelaySeqNum = inst->seqNum; cpu->squashFromMemStall(inst, tid); // Switch On Cache Miss @@ -1038,9 +1031,9 @@ PipelineStage::processInstSchedule(DynInstPtr inst,int &reqs_processed) res_stage_num = inst->nextResStage(); } } else { - DPRINTF(InOrderStage, "[tid:%u]: Instruction [sn:%i] with PC %#x " + DPRINTF(InOrderStage, "[tid:%u]: Instruction [sn:%i] with PC %s " " needed no resources in stage %i.\n", - tid, inst->seqNum, inst->readPC(), stageNum); + tid, inst->seqNum, inst->pcState(), stageNum); } return last_req_completed; @@ -1134,8 +1127,8 @@ PipelineStage::dumpInsts() while (!copy_buff.empty()) { DynInstPtr inst = copy_buff.front(); - cprintf("Inst. PC:%#x\n[tid:%i]\n[sn:%i]\n\n", - inst->readPC(), inst->threadNumber, inst->seqNum); + cprintf("Inst. PC:%s\n[tid:%i]\n[sn:%i]\n\n", + inst->pcState(), inst->threadNumber, inst->seqNum); copy_buff.pop(); } diff --git a/src/cpu/inorder/pipeline_stage.hh b/src/cpu/inorder/pipeline_stage.hh index c971e400e..6df104e6c 100644 --- a/src/cpu/inorder/pipeline_stage.hh +++ b/src/cpu/inorder/pipeline_stage.hh @@ -293,15 +293,15 @@ class PipelineStage /** SeqNum of Squashing Branch Delay Instruction (used for MIPS) */ Addr bdelayDoneSeqNum[ThePipeline::MaxThreads]; - /** Instruction used for squashing branch (used for MIPS) */ - DynInstPtr squashInst[ThePipeline::MaxThreads]; - /** Tells when their is a pending delay slot inst. to send * to rename. If there is, then wait squash after the next * instruction (used for MIPS). */ bool squashAfterDelaySlot[ThePipeline::MaxThreads]; + /** Instruction used for squashing branch (used for MIPS) */ + DynInstPtr squashInst[ThePipeline::MaxThreads]; + /** Maximum size of the inter-stage buffer connecting the previous stage to * this stage (which we call a skid buffer) */ unsigned stageBufferMax; diff --git a/src/cpu/inorder/resources/bpred_unit.cc b/src/cpu/inorder/resources/bpred_unit.cc index b08a393f7..310053409 100644 --- a/src/cpu/inorder/resources/bpred_unit.cc +++ b/src/cpu/inorder/resources/bpred_unit.cc @@ -31,6 +31,7 @@ #include <list> #include <vector> +#include "arch/utility.hh" #include "base/trace.hh" #include "base/traceflags.hh" #include "config/the_isa.hh" @@ -149,7 +150,7 @@ BPredUnit::takeOverFrom() bool -BPredUnit::predict(DynInstPtr &inst, Addr &pred_PC, ThreadID tid) +BPredUnit::predict(DynInstPtr &inst, TheISA::PCState &predPC, ThreadID tid) { // See if branch predictor predicts taken. // If so, get its target addr either from the BTB or the RAS. @@ -160,12 +161,13 @@ BPredUnit::predict(DynInstPtr &inst, Addr &pred_PC, ThreadID tid) int asid = inst->asid; bool pred_taken = false; - Addr target; + TheISA::PCState target; ++lookups; - DPRINTF(InOrderBPred, "[tid:%i] [sn:%i] %s ... PC%#x doing branch " + DPRINTF(InOrderBPred, "[tid:%i] [sn:%i] %s ... PC %s doing branch " "prediction\n", tid, inst->seqNum, - inst->staticInst->disassemble(inst->PC), inst->readPC()); + inst->staticInst->disassemble(inst->instAddr()), + inst->pcState()); void *bp_history = NULL; @@ -185,14 +187,14 @@ BPredUnit::predict(DynInstPtr &inst, Addr &pred_PC, ThreadID tid) } else { ++condPredicted; - pred_taken = BPLookup(pred_PC, bp_history); + pred_taken = BPLookup(predPC.instAddr(), bp_history); DPRINTF(InOrderBPred, "[tid:%i]: Branch predictor predicted %i " - "for PC %#x\n", - tid, pred_taken, inst->readPC()); + "for PC %s\n", + tid, pred_taken, inst->pcState()); } - PredictorHistory predict_record(inst->seqNum, pred_PC, pred_taken, + PredictorHistory predict_record(inst->seqNum, predPC, pred_taken, bp_history, tid); // Now lookup in the BTB or RAS. @@ -202,40 +204,37 @@ BPredUnit::predict(DynInstPtr &inst, Addr &pred_PC, ThreadID tid) // If it's a function return call, then look up the address // in the RAS. - target = RAS[tid].top(); + TheISA::PCState rasTop = RAS[tid].top(); + target = TheISA::buildRetPC(inst->pcState(), rasTop); // Record the top entry of the RAS, and its index. predict_record.usedRAS = true; predict_record.RASIndex = RAS[tid].topIdx(); - predict_record.RASTarget = target; + predict_record.rasTarget = rasTop; assert(predict_record.RASIndex < 16); RAS[tid].pop(); - DPRINTF(InOrderBPred, "[tid:%i]: Instruction %#x is a return, " - "RAS predicted target: %#x, RAS index: %i.\n", - tid, inst->readPC(), target, predict_record.RASIndex); + DPRINTF(InOrderBPred, "[tid:%i]: Instruction %s is a return, " + "RAS predicted target: %s, RAS index: %i.\n", + tid, inst->pcState(), target, + predict_record.RASIndex); } else { ++BTBLookups; if (inst->isCall()) { -#if ISA_HAS_DELAY_SLOT - Addr ras_pc = pred_PC + instSize; // Next Next PC -#else - Addr ras_pc = pred_PC; // Next PC -#endif - - RAS[tid].push(ras_pc); + RAS[tid].push(inst->pcState()); // Record that it was a call so that the top RAS entry can // be popped off if the speculation is incorrect. predict_record.wasCall = true; - DPRINTF(InOrderBPred, "[tid:%i]: Instruction %#x was a call" - ", adding %#x to the RAS index: %i.\n", - tid, inst->readPC(), ras_pc, RAS[tid].topIdx()); + DPRINTF(InOrderBPred, "[tid:%i]: Instruction %s was a call" + ", adding %s to the RAS index: %i.\n", + tid, inst->pcState(), predPC, + RAS[tid].topIdx()); } if (inst->isCall() && @@ -243,18 +242,18 @@ BPredUnit::predict(DynInstPtr &inst, Addr &pred_PC, ThreadID tid) inst->isDirectCtrl()) { target = inst->branchTarget(); - DPRINTF(InOrderBPred, "[tid:%i]: Setting %#x predicted" - " target to %#x.\n", - tid, inst->readPC(), target); - } else if (BTB.valid(pred_PC, asid)) { + DPRINTF(InOrderBPred, "[tid:%i]: Setting %s predicted" + " target to %s.\n", + tid, inst->pcState(), target); + } else if (BTB.valid(predPC.instAddr(), asid)) { ++BTBHits; // If it's not a return, use the BTB to get the target addr. - target = BTB.lookup(pred_PC, asid); + target = BTB.lookup(predPC.instAddr(), asid); - DPRINTF(InOrderBPred, "[tid:%i]: [asid:%i] Instruction %#x " - "predicted target is %#x.\n", - tid, asid, inst->readPC(), target); + DPRINTF(InOrderBPred, "[tid:%i]: [asid:%i] Instruction %s " + "predicted target is %s.\n", + tid, asid, inst->pcState(), target); } else { DPRINTF(InOrderBPred, "[tid:%i]: BTB doesn't have a " "valid entry.\n",tid); @@ -265,14 +264,7 @@ BPredUnit::predict(DynInstPtr &inst, Addr &pred_PC, ThreadID tid) if (pred_taken) { // Set the PC and the instruction's predicted target. - pred_PC = target; - } else { -#if ISA_HAS_DELAY_SLOT - // This value will be inst->PC + 4 (nextPC) - // Delay Slot archs need this to be inst->PC + 8 (nextNPC) - // so we increment one more time here. - pred_PC = pred_PC + instSize; -#endif + predPC = target; } predHist[tid].push_front(predict_record); @@ -296,7 +288,7 @@ BPredUnit::update(const InstSeqNum &done_sn, ThreadID tid) while (!predHist[tid].empty() && predHist[tid].back().seqNum <= done_sn) { // Update the branch predictor with the correct results. - BPUpdate(predHist[tid].back().PC, + BPUpdate(predHist[tid].back().pc.instAddr(), predHist[tid].back().predTaken, predHist[tid].back().bpHistory); @@ -314,13 +306,13 @@ BPredUnit::squash(const InstSeqNum &squashed_sn, ThreadID tid, ThreadID asid) pred_hist.front().seqNum > squashed_sn) { if (pred_hist.front().usedRAS) { DPRINTF(InOrderBPred, "BranchPred: [tid:%i]: Restoring top of RAS " - "to: %i, target: %#x.\n", + "to: %i, target: %s.\n", tid, pred_hist.front().RASIndex, - pred_hist.front().RASTarget); + pred_hist.front().rasTarget); RAS[tid].restore(pred_hist.front().RASIndex, - pred_hist.front().RASTarget); + pred_hist.front().rasTarget); } else if (pred_hist.front().wasCall) { DPRINTF(InOrderBPred, "BranchPred: [tid:%i]: Removing speculative " @@ -340,7 +332,7 @@ BPredUnit::squash(const InstSeqNum &squashed_sn, ThreadID tid, ThreadID asid) void BPredUnit::squash(const InstSeqNum &squashed_sn, - const Addr &corr_target, + const TheISA::PCState &corrTarget, bool actually_taken, ThreadID tid, ThreadID asid) @@ -354,8 +346,8 @@ BPredUnit::squash(const InstSeqNum &squashed_sn, ++condIncorrect; DPRINTF(InOrderBPred, "[tid:%i]: Squashing from sequence number %i, " - "setting target to %#x.\n", - tid, squashed_sn, corr_target); + "setting target to %s.\n", + tid, squashed_sn, corrTarget); squash(squashed_sn, tid); @@ -380,13 +372,13 @@ BPredUnit::squash(const InstSeqNum &squashed_sn, ++RASIncorrect; } - BPUpdate((*hist_it).PC, actually_taken, + BPUpdate((*hist_it).pc.instAddr(), actually_taken, pred_hist.front().bpHistory); - BTB.update((*hist_it).PC, corr_target, asid); + BTB.update((*hist_it).pc.instAddr(), corrTarget, asid); DPRINTF(InOrderBPred, "[tid:%i]: Removing history for [sn:%i] " - "PC %#x.\n", tid, (*hist_it).seqNum, (*hist_it).PC); + "PC %s.\n", tid, (*hist_it).seqNum, (*hist_it).pc); pred_hist.erase(hist_it); @@ -424,7 +416,7 @@ BPredUnit::BPSquash(void *bp_history) bool -BPredUnit::BPLookup(Addr &inst_PC, void * &bp_history) +BPredUnit::BPLookup(Addr inst_PC, void * &bp_history) { if (predictor == Local) { return localBP->lookup(inst_PC, bp_history); @@ -437,7 +429,7 @@ BPredUnit::BPLookup(Addr &inst_PC, void * &bp_history) void -BPredUnit::BPUpdate(Addr &inst_PC, bool taken, void *bp_history) +BPredUnit::BPUpdate(Addr inst_PC, bool taken, void *bp_history) { if (predictor == Local) { localBP->update(inst_PC, taken, bp_history); diff --git a/src/cpu/inorder/resources/bpred_unit.hh b/src/cpu/inorder/resources/bpred_unit.hh index 881bfcc95..3b1c0f4ef 100644 --- a/src/cpu/inorder/resources/bpred_unit.hh +++ b/src/cpu/inorder/resources/bpred_unit.hh @@ -83,11 +83,12 @@ class BPredUnit * Predicts whether or not the instruction is a taken branch, and the * target of the branch if it is taken. * @param inst The branch instruction. - * @param pred_PC The predicted PC is passed back through this parameter. + * @param predPC The predicted PC is passed back through this parameter. * @param tid The thread id. * @return Returns if the branch is taken or not. */ - bool predict(ThePipeline::DynInstPtr &inst, Addr &pred_PC, ThreadID tid); + bool predict(ThePipeline::DynInstPtr &inst, + TheISA::PCState &predPC, ThreadID tid); // @todo: Rename this function. void BPUncond(void * &bp_history); @@ -114,12 +115,13 @@ class BPredUnit * corrects that sn's update with the proper address and taken/not taken. * @param squashed_sn The sequence number to squash any younger updates up * until. - * @param corr_target The correct branch target. + * @param corrTarget The correct branch target. * @param actually_taken The correct branch direction. * @param tid The thread id. */ - void squash(const InstSeqNum &squashed_sn, const Addr &corr_target, - bool actually_taken, ThreadID tid, ThreadID asid = 0); + void squash(const InstSeqNum &squashed_sn, + const TheISA::PCState &corrTarget, bool actually_taken, + ThreadID tid, ThreadID asid = 0); /** * @param bp_history Pointer to the history object. The predictor @@ -134,7 +136,7 @@ class BPredUnit * has the branch predictor state associated with the lookup. * @return Whether the branch is taken or not taken. */ - bool BPLookup(Addr &inst_PC, void * &bp_history); + bool BPLookup(Addr instPC, void * &bp_history); /** * Looks up a given PC in the BTB to see if a matching entry exists. @@ -149,26 +151,26 @@ class BPredUnit * @param inst_PC The PC to look up. * @return The address of the target of the branch. */ - Addr BTBLookup(Addr &inst_PC) - { return BTB.lookup(inst_PC, 0); } + TheISA::PCState BTBLookup(Addr instPC) + { return BTB.lookup(instPC, 0); } /** * Updates the BP with taken/not taken information. - * @param inst_PC The branch's PC that will be updated. + * @param instPC The branch's PC that will be updated. * @param taken Whether the branch was taken or not taken. * @param bp_history Pointer to the branch predictor state that is * associated with the branch lookup that is being updated. * @todo Make this update flexible enough to handle a global predictor. */ - void BPUpdate(Addr &inst_PC, bool taken, void *bp_history); + void BPUpdate(Addr instPC, bool taken, void *bp_history); /** * Updates the BTB with the target of a branch. * @param inst_PC The branch's PC that will be updated. * @param target_PC The branch's target that will be added to the BTB. */ - void BTBUpdate(Addr &inst_PC, Addr &target_PC) - { BTB.update(inst_PC, target_PC,0); } + void BTBUpdate(Addr instPC, const TheISA::PCState &targetPC) + { BTB.update(instPC, targetPC, 0); } void dump(); @@ -181,22 +183,22 @@ class BPredUnit * Makes a predictor history struct that contains any * information needed to update the predictor, BTB, and RAS. */ - PredictorHistory(const InstSeqNum &seq_num, const Addr &inst_PC, - bool pred_taken, void *bp_history, - ThreadID _tid) - : seqNum(seq_num), PC(inst_PC), RASTarget(0), - RASIndex(0), tid(_tid), predTaken(pred_taken), usedRAS(0), - wasCall(0), bpHistory(bp_history) - { } + PredictorHistory(const InstSeqNum &seq_num, + const TheISA::PCState &instPC, bool pred_taken, + void *bp_history, ThreadID _tid) + : seqNum(seq_num), pc(instPC), rasTarget(0), RASIndex(0), + tid(_tid), predTaken(pred_taken), usedRAS(0), wasCall(0), + bpHistory(bp_history) + {} /** The sequence number for the predictor history entry. */ InstSeqNum seqNum; /** The PC associated with the sequence number. */ - Addr PC; + TheISA::PCState pc; /** The RAS target (only valid if a return). */ - Addr RASTarget; + TheISA::PCState rasTarget; /** The RAS index of the instruction (only valid if a call). */ unsigned RASIndex; diff --git a/src/cpu/inorder/resources/branch_predictor.cc b/src/cpu/inorder/resources/branch_predictor.cc index b971d959a..33b67ce4a 100644 --- a/src/cpu/inorder/resources/branch_predictor.cc +++ b/src/cpu/inorder/resources/branch_predictor.cc @@ -84,41 +84,34 @@ BranchPredictor::execute(int slot_num) DPRINTF(InOrderStage, "[tid:%u]: [sn:%i]: squashed, " "skipping prediction \n", tid, inst->seqNum); } else { - Addr pred_PC = inst->readNextPC(); + TheISA::PCState predPC = inst->pcState(); + TheISA::advancePC(predPC, inst->staticInst); if (inst->isControl()) { // If not, the pred_PC be updated to pc+8 // If predicted, the pred_PC will be updated to new target // value - bool predict_taken = branchPred.predict(inst, pred_PC, tid); + bool predict_taken = branchPred.predict(inst, predPC, tid); if (predict_taken) { DPRINTF(InOrderBPred, "[tid:%i]: [sn:%i]: Branch " "predicted true.\n", tid, seq_num); - - inst->setPredTarg(pred_PC); - predictedTaken++; } else { DPRINTF(InOrderBPred, "[tid:%i]: [sn:%i]: Branch " "predicted false.\n", tid, seq_num); - - if (inst->isCondDelaySlot()) - { - inst->setPredTarg(inst->readPC() + (2 * instSize)); - } else { - inst->setPredTarg(pred_PC); - } - predictedNotTaken++; } + inst->setPredTarg(predPC); + inst->setBranchPred(predict_taken); DPRINTF(InOrderBPred, "[tid:%i]: [sn:%i]: Predicted PC is " - "%08p.\n", tid, seq_num, pred_PC); + "%s.\n", tid, seq_num, predPC); } else { + inst->setPredTarg(predPC); //DPRINTF(InOrderBPred, "[tid:%i]: Ignoring [sn:%i] " // "because this isn't " // "a control instruction.\n", tid, seq_num); @@ -166,10 +159,9 @@ BranchPredictor::squash(DynInstPtr inst, int squash_stage, squash_seq_num = squash_seq_num - 1; #endif - if(squash_stage>=ThePipeline::BackEndStartStage) { - Addr corr_targ=inst->readPredPC(); - bool taken=inst->predTaken(); - branchPred.squash(squash_seq_num,corr_targ,taken,tid); + if (squash_stage >= ThePipeline::BackEndStartStage) { + bool taken = inst->predTaken(); + branchPred.squash(squash_seq_num, inst->readPredTarg(), taken, tid); } else { branchPred.squash(squash_seq_num, tid); } diff --git a/src/cpu/inorder/resources/cache_unit.cc b/src/cpu/inorder/resources/cache_unit.cc index 73deacb12..e7f689ffa 100644 --- a/src/cpu/inorder/resources/cache_unit.cc +++ b/src/cpu/inorder/resources/cache_unit.cc @@ -392,15 +392,16 @@ CacheUnit::doTLBAccess(DynInstPtr inst, CacheReqPtr cache_req, int acc_size, unsigned slot_idx = cache_req->getSlot(); if (tlb_mode == TheISA::TLB::Execute) { - inst->fetchMemReq = new Request(inst->readTid(), aligned_addr, - acc_size, flags, inst->readPC(), - cpu->readCpuId(), inst->readTid()); - cache_req->memReq = inst->fetchMemReq; + inst->fetchMemReq = + new Request(inst->readTid(), aligned_addr, acc_size, flags, + inst->instAddr(), cpu->readCpuId(), inst->readTid()); + cache_req->memReq = inst->fetchMemReq; } else { if (!cache_req->is2ndSplit()) { - inst->dataMemReq = new Request(cpu->asid[tid], aligned_addr, - acc_size, flags, inst->readPC(), - cpu->readCpuId(), inst->readTid()); + inst->dataMemReq = + new Request(cpu->asid[tid], aligned_addr, acc_size, flags, + inst->instAddr(), cpu->readCpuId(), + inst->readTid()); cache_req->memReq = inst->dataMemReq; } else { assert(inst->splitInst); @@ -409,7 +410,7 @@ CacheUnit::doTLBAccess(DynInstPtr inst, CacheReqPtr cache_req, int acc_size, inst->split2ndAddr, acc_size, flags, - inst->readPC(), + inst->instAddr(), cpu->readCpuId(), tid); cache_req->memReq = inst->splitMemReq; @@ -754,7 +755,8 @@ CacheUnit::execute(int slot_num) DPRINTF(InOrderCachePort, "[tid:%i]: Instruction [sn:%i] is: %s\n", - tid, seq_num, inst->staticInst->disassemble(inst->PC)); + tid, seq_num, + inst->staticInst->disassemble(inst->instAddr())); removeAddrDependency(inst); @@ -771,7 +773,7 @@ CacheUnit::execute(int slot_num) tid, inst->seqNum); DPRINTF(InOrderStall, "STALL: [tid:%i]: Fetch miss from %08p\n", - tid, cache_req->inst->readPC()); + tid, cache_req->inst->instAddr()); cache_req->setCompleted(false); //cache_req->setMemStall(true); } @@ -1046,21 +1048,22 @@ CacheUnit::processCacheCompletion(PacketPtr pkt) // @todo: update thsi ExtMachInst ext_inst; StaticInstPtr staticInst = NULL; - Addr inst_pc = inst->readPC(); + TheISA::PCState instPC = inst->pcState(); MachInst mach_inst = TheISA::gtoh(*reinterpret_cast<TheISA::MachInst *> (cache_pkt->getPtr<uint8_t>())); predecoder.setTC(cpu->thread[tid]->getTC()); - predecoder.moreBytes(inst_pc, inst_pc, mach_inst); - ext_inst = predecoder.getExtMachInst(); + predecoder.moreBytes(instPC, inst->instAddr(), mach_inst); + ext_inst = predecoder.getExtMachInst(instPC); + inst->pcState(instPC); inst->setMachInst(ext_inst); // Set Up More TraceData info if (inst->traceData) { inst->traceData->setStaticInst(inst->staticInst); - inst->traceData->setPC(inst->readPC()); + inst->traceData->setPC(instPC); } } else if (inst->staticInst && inst->isMemRef()) { @@ -1149,7 +1152,7 @@ CacheUnit::processCacheCompletion(PacketPtr pkt) } else { DPRINTF(InOrderCachePort, "[tid:%u] Miss on block @ %08p completed, but squashed\n", - tid, cache_req->inst->readPC()); + tid, cache_req->inst->instAddr()); cache_req->setMemAccCompleted(); } } diff --git a/src/cpu/inorder/resources/execution_unit.cc b/src/cpu/inorder/resources/execution_unit.cc index 91e788fbc..4342042e9 100644 --- a/src/cpu/inorder/resources/execution_unit.cc +++ b/src/cpu/inorder/resources/execution_unit.cc @@ -91,8 +91,8 @@ ExecutionUnit::execute(int slot_num) exec_req->fault = NoFault; - DPRINTF(InOrderExecute, "[tid:%i] Executing [sn:%i] [PC:%#x] %s.\n", - tid, seq_num, inst->readPC(), inst->instName()); + DPRINTF(InOrderExecute, "[tid:%i] Executing [sn:%i] [PC:%s] %s.\n", + tid, seq_num, inst->pcState(), inst->instName()); switch (exec_req->cmd) { @@ -124,58 +124,61 @@ ExecutionUnit::execute(int slot_num) if (inst->isDirectCtrl()) { assert(!inst->isIndirectCtrl()); + TheISA::PCState pc = inst->pcState(); + TheISA::advancePC(pc, inst->staticInst); + inst->setPredTarg(pc); + if (inst->predTaken() && inst->isCondDelaySlot()) { inst->bdelaySeqNum = seq_num; - inst->setPredTarg(inst->nextPC); DPRINTF(InOrderExecute, "[tid:%i]: Conditional" - " branch inst [sn:%i] PC %#x mis" + " branch inst [sn:%i] PC %s mis" "predicted as taken.\n", tid, - seq_num, inst->PC); + seq_num, inst->pcState()); } else if (!inst->predTaken() && inst->isCondDelaySlot()) { inst->bdelaySeqNum = seq_num; - inst->setPredTarg(inst->nextPC); inst->procDelaySlotOnMispred = true; DPRINTF(InOrderExecute, "[tid:%i]: Conditional" - " branch inst [sn:%i] PC %#x mis" + " branch inst [sn:%i] PC %s mis" "predicted as not taken.\n", tid, - seq_num, inst->PC); + seq_num, inst->pcState()); } else { #if ISA_HAS_DELAY_SLOT inst->bdelaySeqNum = seq_num + 1; - inst->setPredTarg(inst->nextNPC); #else inst->bdelaySeqNum = seq_num; - inst->setPredTarg(inst->nextPC); #endif DPRINTF(InOrderExecute, "[tid:%i]: " "Misprediction detected at " - "[sn:%i] PC %#x,\n\t squashing after " + "[sn:%i] PC %s,\n\t squashing after " "delay slot instruction [sn:%i].\n", - tid, seq_num, inst->PC, + tid, seq_num, inst->pcState(), inst->bdelaySeqNum); DPRINTF(InOrderStall, "STALL: [tid:%i]: Branch" - " misprediction at %#x\n", - tid, inst->PC); + " misprediction at %s\n", + tid, inst->pcState()); } DPRINTF(InOrderExecute, "[tid:%i] Redirecting " - "fetch to %#x.\n", tid, + "fetch to %s.\n", tid, inst->readPredTarg()); - } else if(inst->isIndirectCtrl()){ + } else if (inst->isIndirectCtrl()){ + TheISA::PCState pc = inst->pcState(); + TheISA::advancePC(pc, inst->staticInst); + inst->seqNum = seq_num; + inst->setPredTarg(pc); + #if ISA_HAS_DELAY_SLOT - inst->setPredTarg(inst->nextNPC); inst->bdelaySeqNum = seq_num + 1; #else - inst->setPredTarg(inst->nextPC); inst->bdelaySeqNum = seq_num; #endif DPRINTF(InOrderExecute, "[tid:%i] Redirecting" - " fetch to %#x.\n", tid, + " fetch to %s.\n", tid, inst->readPredTarg()); } else { panic("Non-control instruction (%s) mispredict" @@ -197,14 +200,20 @@ ExecutionUnit::execute(int slot_num) if (inst->predTaken()) { predictedTakenIncorrect++; - DPRINTF(InOrderExecute, "[tid:%i] [sn:%i] %s ... PC%#x ... Mispredicts! (Taken)\n", - tid, inst->seqNum, inst->staticInst->disassemble(inst->PC), - inst->readPC()); + DPRINTF(InOrderExecute, "[tid:%i] [sn:%i] %s ..." + "PC %s ... Mispredicts! (Taken)\n", + tid, inst->seqNum, + inst->staticInst->disassemble( + inst->instAddr()), + inst->pcState()); } else { predictedNotTakenIncorrect++; - DPRINTF(InOrderExecute, "[tid:%i] [sn:%i] %s ... PC%#x ... Mispredicts! (Not Taken)\n", - tid, inst->seqNum, inst->staticInst->disassemble(inst->PC), - inst->readPC()); + DPRINTF(InOrderExecute, "[tid:%i] [sn:%i] %s ..." + "PC %s ... Mispredicts! (Not Taken)\n", + tid, inst->seqNum, + inst->staticInst->disassemble( + inst->instAddr()), + inst->pcState()); } predictedIncorrect++; } else { diff --git a/src/cpu/inorder/resources/fetch_seq_unit.cc b/src/cpu/inorder/resources/fetch_seq_unit.cc index 8a1ec3ce5..3bfe912e7 100644 --- a/src/cpu/inorder/resources/fetch_seq_unit.cc +++ b/src/cpu/inorder/resources/fetch_seq_unit.cc @@ -44,9 +44,6 @@ FetchSeqUnit::FetchSeqUnit(std::string res_name, int res_id, int res_width, instSize(sizeof(MachInst)) { for (ThreadID tid = 0; tid < ThePipeline::MaxThreads; tid++) { - delaySlotInfo[tid].numInsts = 0; - delaySlotInfo[tid].targetReady = false; - pcValid[tid] = false; pcBlockStage[tid] = 0; @@ -86,53 +83,17 @@ FetchSeqUnit::execute(int slot_num) case AssignNextPC: { if (pcValid[tid]) { + inst->pcState(pc[tid]); + inst->setMemAddr(pc[tid].instAddr()); - if (delaySlotInfo[tid].targetReady && - delaySlotInfo[tid].numInsts == 0) { - // Set PC to target - PC[tid] = delaySlotInfo[tid].targetAddr; //next_PC - nextPC[tid] = PC[tid] + instSize; //next_NPC - nextNPC[tid] = PC[tid] + (2 * instSize); - - delaySlotInfo[tid].targetReady = false; - - DPRINTF(InOrderFetchSeq, "[tid:%i]: Setting PC to delay " - "slot target\n",tid); - } + pc[tid].advance(); //XXX HACK! + inst->setPredTarg(pc[tid]); - inst->setPC(PC[tid]); - inst->setNextPC(PC[tid] + instSize); - inst->setNextNPC(PC[tid] + (instSize * 2)); - -#if ISA_HAS_DELAY_SLOT - inst->setPredTarg(inst->readNextNPC()); -#else - inst->setPredTarg(inst->readNextPC()); -#endif - inst->setMemAddr(PC[tid]); inst->setSeqNum(cpu->getAndIncrementInstSeq(tid)); DPRINTF(InOrderFetchSeq, "[tid:%i]: Assigning [sn:%i] to " - "PC %08p, NPC %08p, NNPC %08p\n", tid, - inst->seqNum, inst->readPC(), inst->readNextPC(), - inst->readNextNPC()); - - if (delaySlotInfo[tid].numInsts > 0) { - --delaySlotInfo[tid].numInsts; - - // It's OK to set PC to target of branch - if (delaySlotInfo[tid].numInsts == 0) { - delaySlotInfo[tid].targetReady = true; - } - - DPRINTF(InOrderFetchSeq, "[tid:%i]: %i delay slot inst(s) " - "left to process.\n", tid, - delaySlotInfo[tid].numInsts); - } - - PC[tid] = nextPC[tid]; - nextPC[tid] = nextNPC[tid]; - nextNPC[tid] += instSize; + "PC %s\n", tid, inst->seqNum, + inst->pcState()); fs_req->done(); } else { @@ -147,18 +108,21 @@ FetchSeqUnit::execute(int slot_num) if (inst->isControl()) { // If it's a return, then we must wait for resolved address. if (inst->isReturn() && !inst->predTaken()) { - cpu->pipelineStage[stage_num]->toPrevStages->stageBlock[stage_num][tid] = true; + cpu->pipelineStage[stage_num]-> + toPrevStages->stageBlock[stage_num][tid] = true; pcValid[tid] = false; pcBlockStage[tid] = stage_num; } else if (inst->isCondDelaySlot() && !inst->predTaken()) { // Not-Taken AND Conditional Control - DPRINTF(InOrderFetchSeq, "[tid:%i]: [sn:%i]: [PC:%08p] " - "Predicted Not-Taken Cond. " - "Delay inst. Skipping delay slot and Updating PC to %08p\n", - tid, inst->seqNum, inst->readPC(), inst->readPredTarg()); + DPRINTF(InOrderFetchSeq, "[tid:%i]: [sn:%i]: [PC:%s] " + "Predicted Not-Taken Cond. Delay inst. Skipping " + "delay slot and Updating PC to %s\n", + tid, inst->seqNum, inst->pcState(), + inst->readPredTarg()); - DPRINTF(InOrderFetchSeq, "[tid:%i] Setting up squash to start from stage %i, after [sn:%i].\n", - tid, stage_num, seq_num); + DPRINTF(InOrderFetchSeq, "[tid:%i] Setting up squash to " + "start from stage %i, after [sn:%i].\n", tid, + stage_num, seq_num); inst->bdelaySeqNum = seq_num; inst->squashingStage = stage_num; @@ -168,33 +132,26 @@ FetchSeqUnit::execute(int slot_num) // Not-Taken Control DPRINTF(InOrderFetchSeq, "[tid:%i]: [sn:%i]: Predicted " "Not-Taken Control " - "inst. updating PC to %08p\n", tid, inst->seqNum, - inst->readNextPC()); + "inst. updating PC to %s\n", tid, inst->seqNum, + inst->readPredTarg()); #if ISA_HAS_DELAY_SLOT - ++delaySlotInfo[tid].numInsts; - delaySlotInfo[tid].targetReady = false; - delaySlotInfo[tid].targetAddr = inst->readNextNPC(); -#else - assert(delaySlotInfo[tid].numInsts == 0); + pc[tid] = inst->pcState(); + advancePC(pc[tid], inst->staticInst); #endif } else if (inst->predTaken()) { // Taken Control #if ISA_HAS_DELAY_SLOT - ++delaySlotInfo[tid].numInsts; - delaySlotInfo[tid].targetReady = false; - delaySlotInfo[tid].targetAddr = inst->readPredTarg(); + pc[tid] = inst->readPredTarg(); DPRINTF(InOrderFetchSeq, "[tid:%i]: [sn:%i] Updating delay" - " slot target to PC %08p\n", tid, inst->seqNum, + " slot target to PC %s\n", tid, inst->seqNum, inst->readPredTarg()); inst->bdelaySeqNum = seq_num + 1; #else inst->bdelaySeqNum = seq_num; - assert(delaySlotInfo[tid].numInsts == 0); #endif inst->squashingStage = stage_num; - DPRINTF(InOrderFetchSeq, "[tid:%i] Setting up squash to " "start from stage %i, after [sn:%i].\n", tid, stage_num, inst->bdelaySeqNum); @@ -225,11 +182,12 @@ FetchSeqUnit::squashAfterInst(DynInstPtr inst, int stage_num, ThreadID tid) // Squash inside current resource, so if there needs to be fetching on // same cycle the fetch information will be correct. - // squash(inst, stage_num, inst->bdelaySeqNum, tid); // Schedule Squash Through-out Resource Pool - cpu->resPool->scheduleEvent((InOrderCPU::CPUEventType)ResourcePool::SquashAll, inst, 0); + cpu->resPool->scheduleEvent( + (InOrderCPU::CPUEventType)ResourcePool::SquashAll, inst, 0); } + void FetchSeqUnit::squash(DynInstPtr inst, int squash_stage, InstSeqNum squash_seq_num, ThreadID tid) @@ -241,8 +199,15 @@ FetchSeqUnit::squash(DynInstPtr inst, int squash_stage, // Handles the case where we are squashing because of something that is // not a branch...like a memory stall - Addr new_PC = (inst->isControl()) ? - inst->readPredTarg() : inst->readPC() + instSize; + TheISA::PCState newPC; + if (inst->isControl()) { + newPC = inst->readPredTarg(); + } else { + TheISA::PCState thisPC = inst->pcState(); + assert(inst->staticInst); + advancePC(thisPC, inst->staticInst); + newPC = thisPC; + } if (squashSeqNum[tid] <= done_seq_num && lastSquashCycle[tid] == curTick) { @@ -258,31 +223,25 @@ FetchSeqUnit::squash(DynInstPtr inst, int squash_stage, // the last done_seq_num then this is the delay slot inst. if (cpu->nextInstSeqNum(tid) != done_seq_num && !inst->procDelaySlotOnMispred) { - delaySlotInfo[tid].numInsts = 0; - delaySlotInfo[tid].targetReady = false; // Reset PC - PC[tid] = new_PC; - nextPC[tid] = new_PC + instSize; - nextNPC[tid] = new_PC + (2 * instSize); + pc[tid] = newPC; +#if ISA_HAS_DELAY_SLOT + TheISA::advancePC(pc[tid], inst->staticInst); +#endif - DPRINTF(InOrderFetchSeq, "[tid:%i]: Setting PC to %08p.\n", - tid, PC[tid]); + DPRINTF(InOrderFetchSeq, "[tid:%i]: Setting PC to %s.\n", + tid, newPC); } else { -#if !ISA_HAS_DELAY_SLOT - assert(0); -#endif + assert(ISA_HAS_DELAY_SLOT); - delaySlotInfo[tid].numInsts = 1; - delaySlotInfo[tid].targetReady = false; - delaySlotInfo[tid].targetAddr = (inst->procDelaySlotOnMispred) ? - inst->branchTarget() : new_PC; + pc[tid] = (inst->procDelaySlotOnMispred) ? + inst->branchTarget() : newPC; // Reset PC to Delay Slot Instruction if (inst->procDelaySlotOnMispred) { - PC[tid] = new_PC; - nextPC[tid] = new_PC + instSize; - nextNPC[tid] = new_PC + (2 * instSize); + // Reset PC + pc[tid] = newPC; } } @@ -309,18 +268,13 @@ FetchSeqUnit::FetchSeqEvent::process() FetchSeqUnit* fs_res = dynamic_cast<FetchSeqUnit*>(resource); assert(fs_res); - for (int i=0; i < MaxThreads; i++) { - fs_res->PC[i] = fs_res->cpu->readPC(i); - fs_res->nextPC[i] = fs_res->cpu->readNextPC(i); - fs_res->nextNPC[i] = fs_res->cpu->readNextNPC(i); - DPRINTF(InOrderFetchSeq, "[tid:%i]: Setting PC:%08p NPC:%08p " - "NNPC:%08p.\n", fs_res->PC[i], fs_res->nextPC[i], - fs_res->nextNPC[i]); + for (int i = 0; i < MaxThreads; i++) { + fs_res->pc[i] = fs_res->cpu->pcState(i); + DPRINTF(InOrderFetchSeq, "[tid:%i]: Setting PC: %s.\n", + fs_res->pc[i]); fs_res->pcValid[i] = true; } - - //cpu->fetchPriorityList.push_back(tid); } @@ -329,22 +283,17 @@ FetchSeqUnit::activateThread(ThreadID tid) { pcValid[tid] = true; - PC[tid] = cpu->readPC(tid); - nextPC[tid] = cpu->readNextPC(tid); - nextNPC[tid] = cpu->readNextNPC(tid); + pc[tid] = cpu->pcState(tid); cpu->fetchPriorityList.push_back(tid); - DPRINTF(InOrderFetchSeq, "[tid:%i]: Reading PC:%08p NPC:%08p " - "NNPC:%08p.\n", tid, PC[tid], nextPC[tid], nextNPC[tid]); + DPRINTF(InOrderFetchSeq, "[tid:%i]: Reading PC: %s.\n", + tid, pc[tid]); } void FetchSeqUnit::deactivateThread(ThreadID tid) { - delaySlotInfo[tid].numInsts = 0; - delaySlotInfo[tid].targetReady = false; - pcValid[tid] = false; pcBlockStage[tid] = 0; @@ -375,18 +324,14 @@ FetchSeqUnit::updateAfterContextSwitch(DynInstPtr inst, ThreadID tid) * switch was right after the branch. Thus, if it's not, then * we are updating incorrectly here */ - assert(cpu->thread[tid]->lastBranchNextPC == inst->readPC()); - - PC[tid] = cpu->thread[tid]->lastBranchNextNPC; - nextPC[tid] = PC[tid] + instSize; - nextNPC[tid] = nextPC[tid] + instSize; + assert(cpu->nextInstAddr(tid) == inst->instAddr()); + pc[tid] = cpu->thread[tid]->lastBranchPC; } else { - PC[tid] = inst->readNextPC(); - nextPC[tid] = inst->readNextNPC(); - nextNPC[tid] = inst->readNextNPC() + instSize; + pc[tid] = inst->pcState(); } + assert(inst->staticInst); + advancePC(pc[tid], inst->staticInst); DPRINTF(InOrderFetchSeq, "[tid:%i]: Updating PCs due to Context Switch." - "Assigning PC:%08p NPC:%08p NNPC:%08p.\n", tid, PC[tid], - nextPC[tid], nextNPC[tid]); + "Assigning PC: %s.\n", tid, pc[tid]); } diff --git a/src/cpu/inorder/resources/fetch_seq_unit.hh b/src/cpu/inorder/resources/fetch_seq_unit.hh index aab462224..a258dc0e5 100644 --- a/src/cpu/inorder/resources/fetch_seq_unit.hh +++ b/src/cpu/inorder/resources/fetch_seq_unit.hh @@ -80,22 +80,7 @@ class FetchSeqUnit : public Resource { bool pcValid[ThePipeline::MaxThreads]; int pcBlockStage[ThePipeline::MaxThreads]; - TheISA::IntReg PC[ThePipeline::MaxThreads]; - TheISA::IntReg nextPC[ThePipeline::MaxThreads]; - TheISA::IntReg nextNPC[ThePipeline::MaxThreads]; - - /** Tracks delay slot information for threads in ISAs which use - * delay slots; - */ - struct DelaySlotInfo { - InstSeqNum delaySlotSeqNum; - InstSeqNum branchSeqNum; - int numInsts; - Addr targetAddr; - bool targetReady; - }; - - DelaySlotInfo delaySlotInfo[ThePipeline::MaxThreads]; + TheISA::PCState pc[ThePipeline::MaxThreads]; /** Squash Seq. Nums*/ InstSeqNum squashSeqNum[ThePipeline::MaxThreads]; diff --git a/src/cpu/inorder/resources/tlb_unit.hh b/src/cpu/inorder/resources/tlb_unit.hh index 5c62c7751..eb1bf55f0 100644 --- a/src/cpu/inorder/resources/tlb_unit.hh +++ b/src/cpu/inorder/resources/tlb_unit.hh @@ -111,8 +111,10 @@ class TLBUnitRequest : public ResourceRequest { aligned_addr = inst->getMemAddr(); req_size = sizeof(TheISA::MachInst); flags = 0; - inst->fetchMemReq = new Request(inst->readTid(), aligned_addr, req_size, - flags, inst->readPC(), res->cpu->readCpuId(), inst->readTid()); + inst->fetchMemReq = new Request(inst->readTid(), aligned_addr, + req_size, flags, inst->instAddr(), + res->cpu->readCpuId(), + inst->readTid()); memReq = inst->fetchMemReq; } else { aligned_addr = inst->getMemAddr();; @@ -123,8 +125,10 @@ class TLBUnitRequest : public ResourceRequest { req_size = 8; } - inst->dataMemReq = new Request(inst->readTid(), aligned_addr, req_size, - flags, inst->readPC(), res->cpu->readCpuId(), inst->readTid()); + inst->dataMemReq = new Request(inst->readTid(), aligned_addr, + req_size, flags, inst->instAddr(), + res->cpu->readCpuId(), + inst->readTid()); memReq = inst->dataMemReq; } } diff --git a/src/cpu/inorder/thread_context.cc b/src/cpu/inorder/thread_context.cc index e6d0af2cc..bbccf626a 100644 --- a/src/cpu/inorder/thread_context.cc +++ b/src/cpu/inorder/thread_context.cc @@ -234,30 +234,6 @@ InOrderThreadContext::setRegOtherThread(int misc_reg, const MiscReg &val, } void -InOrderThreadContext::setPC(uint64_t val) -{ - DPRINTF(InOrderCPU, "[tid:%i] Setting PC to %08p\n", - thread->readTid(), val); - cpu->setPC(val, thread->readTid()); -} - -void -InOrderThreadContext::setNextPC(uint64_t val) -{ - DPRINTF(InOrderCPU, "[tid:%i] Setting NPC to %08p\n", - thread->readTid(), val); - cpu->setNextPC(val, thread->readTid()); -} - -void -InOrderThreadContext::setNextNPC(uint64_t val) -{ - DPRINTF(InOrderCPU, "[tid:%i] Setting NNPC to %08p\n", - thread->readTid(), val); - cpu->setNextNPC(val, thread->readTid()); -} - -void InOrderThreadContext::setMiscRegNoEffect(int misc_reg, const MiscReg &val) { cpu->setMiscRegNoEffect(misc_reg, val, thread->readTid()); diff --git a/src/cpu/inorder/thread_context.hh b/src/cpu/inorder/thread_context.hh index 0c0f563c0..21f1e5835 100644 --- a/src/cpu/inorder/thread_context.hh +++ b/src/cpu/inorder/thread_context.hh @@ -204,23 +204,21 @@ class InOrderThreadContext : public ThreadContext ThreadID tid); /** Reads this thread's PC. */ - uint64_t readPC() - { return cpu->readPC(thread->readTid()); } + TheISA::PCState pcState() + { return cpu->pcState(thread->readTid()); } /** Sets this thread's PC. */ - void setPC(uint64_t val); + void pcState(const TheISA::PCState &val) + { cpu->pcState(val, thread->readTid()); } - /** Reads this thread's next PC. */ - uint64_t readNextPC() - { return cpu->readNextPC(thread->readTid()); } + Addr instAddr() + { return cpu->instAddr(thread->readTid()); } - /** Sets this thread's next PC. */ - void setNextPC(uint64_t val); + Addr nextInstAddr() + { return cpu->nextInstAddr(thread->readTid()); } - uint64_t readNextNPC() - { return cpu->readNextNPC(thread->readTid()); } - - void setNextNPC(uint64_t val); + MicroPC microPC() + { return cpu->microPC(thread->readTid()); } /** Reads a miscellaneous register. */ MiscReg readMiscRegNoEffect(int misc_reg) diff --git a/src/cpu/inorder/thread_state.hh b/src/cpu/inorder/thread_state.hh index 0a171a99f..20ace6659 100644 --- a/src/cpu/inorder/thread_state.hh +++ b/src/cpu/inorder/thread_state.hh @@ -111,9 +111,7 @@ class InOrderThreadState : public ThreadState { /** Is last instruction graduated a branch? */ bool lastGradIsBranch; - Addr lastBranchPC; - Addr lastBranchNextPC; - Addr lastBranchNextNPC; + TheISA::PCState lastBranchPC; }; #endif // __CPU_INORDER_THREAD_STATE_HH__ diff --git a/src/cpu/inteltrace.cc b/src/cpu/inteltrace.cc index ec51b80e7..ee148c50f 100644 --- a/src/cpu/inteltrace.cc +++ b/src/cpu/inteltrace.cc @@ -48,7 +48,7 @@ Trace::IntelTraceRecord::dump() { ostream &outs = Trace::output(); ccprintf(outs, "%7d ) ", when); - outs << "0x" << hex << PC << ":\t"; + outs << "0x" << hex << pc.instAddr() << ":\t"; if (staticInst->isLoad()) { ccprintf(outs, "<RD %#x>", addr); } else if (staticInst->isStore()) { diff --git a/src/cpu/inteltrace.hh b/src/cpu/inteltrace.hh index c4ace4f4b..234b173e9 100644 --- a/src/cpu/inteltrace.hh +++ b/src/cpu/inteltrace.hh @@ -46,10 +46,10 @@ class IntelTraceRecord : public InstRecord { public: IntelTraceRecord(Tick _when, ThreadContext *_thread, - const StaticInstPtr _staticInst, Addr _pc, bool spec, - const StaticInstPtr _macroStaticInst = NULL, MicroPC _upc = 0) + const StaticInstPtr _staticInst, TheISA::PCState _pc, + bool spec, const StaticInstPtr _macroStaticInst = NULL) : InstRecord(_when, _thread, _staticInst, _pc, spec, - _macroStaticInst, _upc) + _macroStaticInst) { } @@ -65,8 +65,8 @@ class IntelTrace : public InstTracer IntelTraceRecord * getInstRecord(Tick when, ThreadContext *tc, - const StaticInstPtr staticInst, Addr pc, - const StaticInstPtr macroStaticInst = NULL, MicroPC upc = 0) + const StaticInstPtr staticInst, TheISA::PCState pc, + const StaticInstPtr macroStaticInst = NULL) { if (!IsOn(ExecEnable)) return NULL; @@ -78,7 +78,7 @@ class IntelTrace : public InstTracer return NULL; return new IntelTraceRecord(when, tc, - staticInst, pc, tc->misspeculating(), macroStaticInst, upc); + staticInst, pc, tc->misspeculating(), macroStaticInst); } }; diff --git a/src/cpu/legiontrace.cc b/src/cpu/legiontrace.cc index 1390d0807..8750e56e5 100644 --- a/src/cpu/legiontrace.cc +++ b/src/cpu/legiontrace.cc @@ -211,7 +211,7 @@ Trace::LegionTraceRecord::dump() if(!staticInst->isMicroop() || staticInst->isLastMicroop()) { while (!compared) { if (shared_data->flags == OWN_M5) { - m5Pc = PC & SparcISA::PAddrImplMask; + m5Pc = pc.instAddr() & SparcISA::PAddrImplMask; if (bits(shared_data->pstate,3,3)) { m5Pc &= mask(32); } @@ -432,13 +432,14 @@ Trace::LegionTraceRecord::dump() << endl; predecoder.setTC(thread); - predecoder.moreBytes(m5Pc, m5Pc, - shared_data->instruction); + predecoder.moreBytes(m5Pc, m5Pc, shared_data->instruction); assert(predecoder.extMachInstReady()); + PCState tempPC = pc; StaticInstPtr legionInst = - StaticInst::decode(predecoder.getExtMachInst(), lgnPc); + StaticInst::decode(predecoder.getExtMachInst(tempPC), + lgnPc); outs << setfill(' ') << setw(15) << " Legion Inst: " << "0x" << setw(8) << setfill('0') << hex diff --git a/src/cpu/legiontrace.hh b/src/cpu/legiontrace.hh index 829941d4b..a60b9ad10 100644 --- a/src/cpu/legiontrace.hh +++ b/src/cpu/legiontrace.hh @@ -46,10 +46,10 @@ class LegionTraceRecord : public InstRecord { public: LegionTraceRecord(Tick _when, ThreadContext *_thread, - const StaticInstPtr _staticInst, Addr _pc, bool spec, - const StaticInstPtr _macroStaticInst = NULL, MicroPC _upc = 0) + const StaticInstPtr _staticInst, TheISA::PCState _pc, + bool spec, const StaticInstPtr _macroStaticInst = NULL) : InstRecord(_when, _thread, _staticInst, _pc, spec, - _macroStaticInst, _upc) + _macroStaticInst) { } @@ -65,14 +65,14 @@ class LegionTrace : public InstTracer LegionTraceRecord * getInstRecord(Tick when, ThreadContext *tc, - const StaticInstPtr staticInst, Addr pc, - const StaticInstPtr macroStaticInst = NULL, MicroPC upc = 0) + const StaticInstPtr staticInst, TheISA::PCState pc, + const StaticInstPtr macroStaticInst = NULL) { if (tc->misspeculating()) return NULL; return new LegionTraceRecord(when, tc, - staticInst, pc, tc->misspeculating(), macroStaticInst, upc); + staticInst, pc, tc->misspeculating(), macroStaticInst); } }; diff --git a/src/cpu/nativetrace.hh b/src/cpu/nativetrace.hh index 6ad6b0242..5c5b9a66d 100644 --- a/src/cpu/nativetrace.hh +++ b/src/cpu/nativetrace.hh @@ -54,10 +54,10 @@ class NativeTraceRecord : public ExeTracerRecord public: NativeTraceRecord(NativeTrace * _parent, Tick _when, ThreadContext *_thread, - const StaticInstPtr _staticInst, Addr _pc, bool spec, - const StaticInstPtr _macroStaticInst = NULL, MicroPC _upc = 0) + const StaticInstPtr _staticInst, TheISA::PCState _pc, + bool spec, const StaticInstPtr _macroStaticInst = NULL) : ExeTracerRecord(_when, _thread, _staticInst, _pc, spec, - _macroStaticInst, _upc), + _macroStaticInst), parent(_parent) { } @@ -79,14 +79,14 @@ class NativeTrace : public ExeTracer NativeTraceRecord * getInstRecord(Tick when, ThreadContext *tc, - const StaticInstPtr staticInst, Addr pc, - const StaticInstPtr macroStaticInst = NULL, MicroPC upc = 0) + const StaticInstPtr staticInst, TheISA::PCState pc, + const StaticInstPtr macroStaticInst = NULL) { if (tc->misspeculating()) return NULL; return new NativeTraceRecord(this, when, tc, - staticInst, pc, tc->misspeculating(), macroStaticInst, upc); + staticInst, pc, tc->misspeculating(), macroStaticInst); } template<class T> diff --git a/src/cpu/o3/bpred_unit.hh b/src/cpu/o3/bpred_unit.hh index f199bdd75..58b1147c9 100644 --- a/src/cpu/o3/bpred_unit.hh +++ b/src/cpu/o3/bpred_unit.hh @@ -88,7 +88,7 @@ class BPredUnit * @param tid The thread id. * @return Returns if the branch is taken or not. */ - bool predict(DynInstPtr &inst, Addr &PC, ThreadID tid); + bool predict(DynInstPtr &inst, TheISA::PCState &pc, ThreadID tid); // @todo: Rename this function. void BPUncond(void * &bp_history); @@ -118,7 +118,8 @@ class BPredUnit * @param actually_taken The correct branch direction. * @param tid The thread id. */ - void squash(const InstSeqNum &squashed_sn, const Addr &corr_target, + void squash(const InstSeqNum &squashed_sn, + const TheISA::PCState &corr_target, bool actually_taken, ThreadID tid); /** @@ -134,23 +135,23 @@ class BPredUnit * has the branch predictor state associated with the lookup. * @return Whether the branch is taken or not taken. */ - bool BPLookup(Addr &inst_PC, void * &bp_history); + bool BPLookup(Addr instPC, void * &bp_history); /** * Looks up a given PC in the BTB to see if a matching entry exists. * @param inst_PC The PC to look up. * @return Whether the BTB contains the given PC. */ - bool BTBValid(Addr &inst_PC) - { return BTB.valid(inst_PC, 0); } + bool BTBValid(Addr instPC) + { return BTB.valid(instPC, 0); } /** * Looks up a given PC in the BTB to get the predicted target. * @param inst_PC The PC to look up. * @return The address of the target of the branch. */ - Addr BTBLookup(Addr &inst_PC) - { return BTB.lookup(inst_PC, 0); } + TheISA::PCState BTBLookup(Addr instPC) + { return BTB.lookup(instPC, 0); } /** * Updates the BP with taken/not taken information. @@ -160,15 +161,15 @@ class BPredUnit * associated with the branch lookup that is being updated. * @todo Make this update flexible enough to handle a global predictor. */ - void BPUpdate(Addr &inst_PC, bool taken, void *bp_history); + void BPUpdate(Addr instPC, bool taken, void *bp_history); /** * Updates the BTB with the target of a branch. * @param inst_PC The branch's PC that will be updated. * @param target_PC The branch's target that will be added to the BTB. */ - void BTBUpdate(Addr &inst_PC, Addr &target_PC) - { BTB.update(inst_PC, target_PC,0); } + void BTBUpdate(Addr instPC, const TheISA::PCState &target) + { BTB.update(instPC, target, 0); } void dump(); @@ -178,13 +179,13 @@ class BPredUnit * Makes a predictor history struct that contains any * information needed to update the predictor, BTB, and RAS. */ - PredictorHistory(const InstSeqNum &seq_num, const Addr &inst_PC, + PredictorHistory(const InstSeqNum &seq_num, Addr instPC, bool pred_taken, void *bp_history, ThreadID _tid) - : seqNum(seq_num), PC(inst_PC), RASTarget(0), - RASIndex(0), tid(_tid), predTaken(pred_taken), usedRAS(0), + : seqNum(seq_num), pc(instPC), RASTarget(0), RASIndex(0), + tid(_tid), predTaken(pred_taken), usedRAS(0), wasCall(0), bpHistory(bp_history) - { } + {} bool operator==(const PredictorHistory &entry) const { return this->seqNum == entry.seqNum; @@ -194,10 +195,10 @@ class BPredUnit InstSeqNum seqNum; /** The PC associated with the sequence number. */ - Addr PC; + Addr pc; /** The RAS target (only valid if a return). */ - Addr RASTarget; + TheISA::PCState RASTarget; /** The RAS index of the instruction (only valid if a call). */ unsigned RASIndex; diff --git a/src/cpu/o3/bpred_unit_impl.hh b/src/cpu/o3/bpred_unit_impl.hh index ed3471761..14d47df9f 100644 --- a/src/cpu/o3/bpred_unit_impl.hh +++ b/src/cpu/o3/bpred_unit_impl.hh @@ -31,6 +31,7 @@ #include <algorithm> #include "arch/types.hh" +#include "arch/utility.hh" #include "arch/isa_traits.hh" #include "base/trace.hh" #include "base/traceflags.hh" @@ -144,17 +145,15 @@ BPredUnit<Impl>::takeOverFrom() template <class Impl> bool -BPredUnit<Impl>::predict(DynInstPtr &inst, Addr &PC, ThreadID tid) +BPredUnit<Impl>::predict(DynInstPtr &inst, TheISA::PCState &pc, ThreadID tid) { // See if branch predictor predicts taken. // If so, get its target addr either from the BTB or the RAS. // Save off record of branch stuff so the RAS can be fixed // up once it's done. - using TheISA::MachInst; - bool pred_taken = false; - Addr target = PC; + TheISA::PCState target = pc; ++lookups; @@ -168,19 +167,19 @@ BPredUnit<Impl>::predict(DynInstPtr &inst, Addr &PC, ThreadID tid) } else { ++condPredicted; - pred_taken = BPLookup(PC, bp_history); + pred_taken = BPLookup(pc.instAddr(), bp_history); DPRINTF(Fetch, "BranchPred: [tid:%i]: Branch predictor predicted %i " - "for PC %#x\n", - tid, pred_taken, inst->readPC()); + "for PC %s\n", + tid, pred_taken, inst->pcState()); } DPRINTF(Fetch, "BranchPred: [tid:%i]: [sn:%i] Creating prediction history " - "for PC %#x\n", - tid, inst->seqNum, inst->readPC()); + "for PC %s\n", + tid, inst->seqNum, inst->pcState()); - PredictorHistory predict_record(inst->seqNum, PC, pred_taken, - bp_history, tid); + PredictorHistory predict_record(inst->seqNum, pc.instAddr(), + pred_taken, bp_history, tid); // Now lookup in the BTB or RAS. if (pred_taken) { @@ -189,60 +188,58 @@ BPredUnit<Impl>::predict(DynInstPtr &inst, Addr &PC, ThreadID tid) // If it's a function return call, then look up the address // in the RAS. - target = RAS[tid].top(); + TheISA::PCState rasTop = RAS[tid].top(); + target = TheISA::buildRetPC(pc, rasTop); // Record the top entry of the RAS, and its index. predict_record.usedRAS = true; predict_record.RASIndex = RAS[tid].topIdx(); - predict_record.RASTarget = target; + predict_record.RASTarget = rasTop; assert(predict_record.RASIndex < 16); RAS[tid].pop(); - DPRINTF(Fetch, "BranchPred: [tid:%i]: Instruction %#x is a return, " - "RAS predicted target: %#x, RAS index: %i.\n", - tid, inst->readPC(), target, predict_record.RASIndex); + DPRINTF(Fetch, "BranchPred: [tid:%i]: Instruction %s is a return, " + "RAS predicted target: %s, RAS index: %i.\n", + tid, inst->pcState(), target, predict_record.RASIndex); } else { ++BTBLookups; if (inst->isCall()) { -#if ISA_HAS_DELAY_SLOT - Addr ras_pc = PC + (2 * sizeof(MachInst)); // Next Next PC -#else - Addr ras_pc = PC + sizeof(MachInst); // Next PC -#endif - RAS[tid].push(ras_pc); + RAS[tid].push(pc); // Record that it was a call so that the top RAS entry can // be popped off if the speculation is incorrect. predict_record.wasCall = true; - DPRINTF(Fetch, "BranchPred: [tid:%i]: Instruction %#x was a call" - ", adding %#x to the RAS index: %i.\n", - tid, inst->readPC(), ras_pc, RAS[tid].topIdx()); + DPRINTF(Fetch, "BranchPred: [tid:%i]: Instruction %s was a " + "call, adding %s to the RAS index: %i.\n", + tid, inst->pcState(), pc, RAS[tid].topIdx()); } - if (BTB.valid(PC, tid)) { + if (BTB.valid(pc.instAddr(), tid)) { ++BTBHits; // If it's not a return, use the BTB to get the target addr. - target = BTB.lookup(PC, tid); + target = BTB.lookup(pc.instAddr(), tid); - DPRINTF(Fetch, "BranchPred: [tid:%i]: Instruction %#x predicted" - " target is %#x.\n", - tid, inst->readPC(), target); + DPRINTF(Fetch, "BranchPred: [tid:%i]: Instruction %s predicted" + " target is %s.\n", tid, inst->pcState(), target); } else { DPRINTF(Fetch, "BranchPred: [tid:%i]: BTB doesn't have a " "valid entry.\n",tid); pred_taken = false; + TheISA::advancePC(target, inst->staticInst); } } + } else { + TheISA::advancePC(target, inst->staticInst); } - PC = target; + pc = target; predHist[tid].push_front(predict_record); @@ -262,7 +259,7 @@ BPredUnit<Impl>::update(const InstSeqNum &done_sn, ThreadID tid) while (!predHist[tid].empty() && predHist[tid].back().seqNum <= done_sn) { // Update the branch predictor with the correct results. - BPUpdate(predHist[tid].back().PC, + BPUpdate(predHist[tid].back().pc, predHist[tid].back().predTaken, predHist[tid].back().bpHistory); @@ -280,10 +277,8 @@ BPredUnit<Impl>::squash(const InstSeqNum &squashed_sn, ThreadID tid) pred_hist.front().seqNum > squashed_sn) { if (pred_hist.front().usedRAS) { DPRINTF(Fetch, "BranchPred: [tid:%i]: Restoring top of RAS to: %i," - " target: %#x.\n", - tid, - pred_hist.front().RASIndex, - pred_hist.front().RASTarget); + " target: %s.\n", tid, + pred_hist.front().RASIndex, pred_hist.front().RASTarget); RAS[tid].restore(pred_hist.front().RASIndex, pred_hist.front().RASTarget); @@ -298,11 +293,13 @@ BPredUnit<Impl>::squash(const InstSeqNum &squashed_sn, ThreadID tid) BPSquash(pred_hist.front().bpHistory); DPRINTF(Fetch, "BranchPred: [tid:%i]: Removing history for [sn:%i] " - "PC %#x.\n", tid, pred_hist.front().seqNum, pred_hist.front().PC); + "PC %s.\n", tid, pred_hist.front().seqNum, + pred_hist.front().pc); pred_hist.pop_front(); - DPRINTF(Fetch, "[tid:%i]: predHist.size(): %i\n", tid, predHist[tid].size()); + DPRINTF(Fetch, "[tid:%i]: predHist.size(): %i\n", + tid, predHist[tid].size()); } } @@ -310,7 +307,7 @@ BPredUnit<Impl>::squash(const InstSeqNum &squashed_sn, ThreadID tid) template <class Impl> void BPredUnit<Impl>::squash(const InstSeqNum &squashed_sn, - const Addr &corr_target, + const TheISA::PCState &corrTarget, bool actually_taken, ThreadID tid) { @@ -330,8 +327,8 @@ BPredUnit<Impl>::squash(const InstSeqNum &squashed_sn, ++condIncorrect; DPRINTF(Fetch, "BranchPred: [tid:%i]: Squashing from sequence number %i, " - "setting target to %#x.\n", - tid, squashed_sn, corr_target); + "setting target to %s.\n", + tid, squashed_sn, corrTarget); // Squash All Branches AFTER this mispredicted branch squash(squashed_sn, tid); @@ -358,13 +355,13 @@ BPredUnit<Impl>::squash(const InstSeqNum &squashed_sn, ++RASIncorrect; } - BPUpdate((*hist_it).PC, actually_taken, + BPUpdate((*hist_it).pc, actually_taken, pred_hist.front().bpHistory); - BTB.update((*hist_it).PC, corr_target, tid); + BTB.update((*hist_it).pc, corrTarget, tid); DPRINTF(Fetch, "BranchPred: [tid:%i]: Removing history for [sn:%i] " - "PC %#x.\n", tid, (*hist_it).seqNum, (*hist_it).PC); + "PC %s.\n", tid, (*hist_it).seqNum, (*hist_it).pc); pred_hist.erase(hist_it); @@ -397,12 +394,12 @@ BPredUnit<Impl>::BPSquash(void *bp_history) template <class Impl> bool -BPredUnit<Impl>::BPLookup(Addr &inst_PC, void * &bp_history) +BPredUnit<Impl>::BPLookup(Addr instPC, void * &bp_history) { if (predictor == Local) { - return localBP->lookup(inst_PC, bp_history); + return localBP->lookup(instPC, bp_history); } else if (predictor == Tournament) { - return tournamentBP->lookup(inst_PC, bp_history); + return tournamentBP->lookup(instPC, bp_history); } else { panic("Predictor type is unexpected value!"); } @@ -410,12 +407,12 @@ BPredUnit<Impl>::BPLookup(Addr &inst_PC, void * &bp_history) template <class Impl> void -BPredUnit<Impl>::BPUpdate(Addr &inst_PC, bool taken, void *bp_history) +BPredUnit<Impl>::BPUpdate(Addr instPC, bool taken, void *bp_history) { if (predictor == Local) { - localBP->update(inst_PC, taken, bp_history); + localBP->update(instPC, taken, bp_history); } else if (predictor == Tournament) { - tournamentBP->update(inst_PC, taken, bp_history); + tournamentBP->update(instPC, taken, bp_history); } else { panic("Predictor type is unexpected value!"); } @@ -436,9 +433,9 @@ BPredUnit<Impl>::dump() while (pred_hist_it != predHist[i].end()) { cprintf("[sn:%lli], PC:%#x, tid:%i, predTaken:%i, " "bpHistory:%#x\n", - (*pred_hist_it).seqNum, (*pred_hist_it).PC, - (*pred_hist_it).tid, (*pred_hist_it).predTaken, - (*pred_hist_it).bpHistory); + pred_hist_it->seqNum, pred_hist_it->pc, + pred_hist_it->tid, pred_hist_it->predTaken, + pred_hist_it->bpHistory); pred_hist_it++; } diff --git a/src/cpu/o3/comm.hh b/src/cpu/o3/comm.hh index 23b836f73..c9fb3319b 100644 --- a/src/cpu/o3/comm.hh +++ b/src/cpu/o3/comm.hh @@ -33,6 +33,7 @@ #include <vector> +#include "arch/types.hh" #include "base/types.hh" #include "cpu/inst_seq.hh" #include "sim/faults.hh" @@ -88,9 +89,7 @@ struct DefaultIEWDefaultCommit { bool branchMispredict[Impl::MaxThreads]; bool branchTaken[Impl::MaxThreads]; Addr mispredPC[Impl::MaxThreads]; - Addr nextPC[Impl::MaxThreads]; - Addr nextNPC[Impl::MaxThreads]; - Addr nextMicroPC[Impl::MaxThreads]; + TheISA::PCState pc[Impl::MaxThreads]; InstSeqNum squashedSeqNum[Impl::MaxThreads]; bool includeSquashInst[Impl::MaxThreads]; @@ -120,9 +119,7 @@ struct TimeBufStruct { bool branchMispredict; bool branchTaken; Addr mispredPC; - Addr nextPC; - Addr nextNPC; - Addr nextMicroPC; + TheISA::PCState nextPC; unsigned branchCount; }; @@ -161,9 +158,7 @@ struct TimeBufStruct { bool branchMispredict; bool branchTaken; Addr mispredPC; - Addr nextPC; - Addr nextNPC; - Addr nextMicroPC; + TheISA::PCState pc; // Represents the instruction that has either been retired or // squashed. Similar to having a single bus that broadcasts the diff --git a/src/cpu/o3/commit.hh b/src/cpu/o3/commit.hh index d93b85984..326f3a1d3 100644 --- a/src/cpu/o3/commit.hh +++ b/src/cpu/o3/commit.hh @@ -277,40 +277,21 @@ class DefaultCommit ThreadID oldestReady(); public: - /** Returns the PC of the head instruction of the ROB. - * @todo: Probably remove this function as it returns only thread 0. - */ - Addr readPC() { return PC[0]; } - - /** Returns the PC of a specific thread. */ - Addr readPC(ThreadID tid) { return PC[tid]; } + /** Reads the PC of a specific thread. */ + TheISA::PCState pcState(ThreadID tid) { return pc[tid]; } /** Sets the PC of a specific thread. */ - void setPC(Addr val, ThreadID tid) { PC[tid] = val; } - - /** Reads the micro PC of a specific thread. */ - Addr readMicroPC(ThreadID tid) { return microPC[tid]; } - - /** Sets the micro PC of a specific thread */ - void setMicroPC(Addr val, ThreadID tid) { microPC[tid] = val; } - - /** Reads the next PC of a specific thread. */ - Addr readNextPC(ThreadID tid) { return nextPC[tid]; } - - /** Sets the next PC of a specific thread. */ - void setNextPC(Addr val, ThreadID tid) { nextPC[tid] = val; } + void pcState(const TheISA::PCState &val, ThreadID tid) + { pc[tid] = val; } - /** Reads the next NPC of a specific thread. */ - Addr readNextNPC(ThreadID tid) { return nextNPC[tid]; } + /** Returns the PC of a specific thread. */ + Addr instAddr(ThreadID tid) { return pc[tid].instAddr(); } - /** Sets the next NPC of a specific thread. */ - void setNextNPC(Addr val, ThreadID tid) { nextNPC[tid] = val; } + /** Returns the next PC of a specific thread. */ + Addr nextInstAddr(ThreadID tid) { return pc[tid].nextInstAddr(); } /** Reads the micro PC of a specific thread. */ - Addr readNextMicroPC(ThreadID tid) { return nextMicroPC[tid]; } - - /** Sets the micro PC of a specific thread */ - void setNextMicroPC(Addr val, ThreadID tid) { nextMicroPC[tid] = val; } + Addr microPC(ThreadID tid) { return pc[tid].microPC(); } private: /** Time buffer interface. */ @@ -410,24 +391,10 @@ class DefaultCommit /** The interrupt fault. */ Fault interrupt; - /** The commit PC of each thread. Refers to the instruction that - * is currently being processed/committed. - */ - Addr PC[Impl::MaxThreads]; - - /** The commit micro PC of each thread. Refers to the instruction that + /** The commit PC state of each thread. Refers to the instruction that * is currently being processed/committed. */ - Addr microPC[Impl::MaxThreads]; - - /** The next PC of each thread. */ - Addr nextPC[Impl::MaxThreads]; - - /** The next NPC of each thread. */ - Addr nextNPC[Impl::MaxThreads]; - - /** The next micro PC of each thread. */ - Addr nextMicroPC[Impl::MaxThreads]; + TheISA::PCState pc[Impl::MaxThreads]; /** The sequence number of the youngest valid instruction in the ROB. */ InstSeqNum youngestSeqNum[Impl::MaxThreads]; diff --git a/src/cpu/o3/commit_impl.hh b/src/cpu/o3/commit_impl.hh index 98c7b49c8..8d3edfb19 100644 --- a/src/cpu/o3/commit_impl.hh +++ b/src/cpu/o3/commit_impl.hh @@ -128,11 +128,7 @@ DefaultCommit<Impl>::DefaultCommit(O3CPU *_cpu, DerivO3CPUParams *params) committedStores[tid] = false; trapSquash[tid] = false; tcSquash[tid] = false; - microPC[tid] = 0; - nextMicroPC[tid] = 0; - PC[tid] = 0; - nextPC[tid] = 0; - nextNPC[tid] = 0; + pc[tid].set(0); } #if FULL_SYSTEM interrupt = NoFault; @@ -513,9 +509,7 @@ DefaultCommit<Impl>::squashAll(ThreadID tid) toIEW->commitInfo[tid].branchMispredict = false; - toIEW->commitInfo[tid].nextPC = PC[tid]; - toIEW->commitInfo[tid].nextNPC = nextPC[tid]; - toIEW->commitInfo[tid].nextMicroPC = nextMicroPC[tid]; + toIEW->commitInfo[tid].pc = pc[tid]; } template <class Impl> @@ -524,7 +518,7 @@ DefaultCommit<Impl>::squashFromTrap(ThreadID tid) { squashAll(tid); - DPRINTF(Commit, "Squashing from trap, restarting at PC %#x\n", PC[tid]); + DPRINTF(Commit, "Squashing from trap, restarting at PC %s\n", pc[tid]); thread[tid]->trapPending = false; thread[tid]->inSyscall = false; @@ -542,7 +536,7 @@ DefaultCommit<Impl>::squashFromTC(ThreadID tid) { squashAll(tid); - DPRINTF(Commit, "Squashing from TC, restarting at PC %#x\n", PC[tid]); + DPRINTF(Commit, "Squashing from TC, restarting at PC %s\n", pc[tid]); thread[tid]->inSyscall = false; assert(!thread[tid]->trapPending); @@ -611,16 +605,16 @@ DefaultCommit<Impl>::tick() DynInstPtr inst = rob->readHeadInst(tid); - DPRINTF(Commit,"[tid:%i]: Instruction [sn:%lli] PC %#x is head of" + DPRINTF(Commit,"[tid:%i]: Instruction [sn:%lli] PC %s is head of" " ROB and ready to commit\n", - tid, inst->seqNum, inst->readPC()); + tid, inst->seqNum, inst->pcState()); } else if (!rob->isEmpty(tid)) { DynInstPtr inst = rob->readHeadInst(tid); DPRINTF(Commit,"[tid:%i]: Can't commit, Instruction [sn:%lli] PC " - "%#x is head of ROB and not ready\n", - tid, inst->seqNum, inst->readPC()); + "%s is head of ROB and not ready\n", + tid, inst->seqNum, inst->pcState()); } DPRINTF(Commit, "[tid:%i]: ROB has %d insts & %d free entries.\n", @@ -738,7 +732,7 @@ DefaultCommit<Impl>::commit() DPRINTF(Commit, "[tid:%i]: Redirecting to PC %#x\n", tid, - fromIEW->nextPC[tid]); + fromIEW->pc[tid].nextInstAddr()); commitStatus[tid] = ROBSquashing; @@ -771,9 +765,7 @@ DefaultCommit<Impl>::commit() toIEW->commitInfo[tid].branchTaken = fromIEW->branchTaken[tid]; - toIEW->commitInfo[tid].nextPC = fromIEW->nextPC[tid]; - toIEW->commitInfo[tid].nextNPC = fromIEW->nextNPC[tid]; - toIEW->commitInfo[tid].nextMicroPC = fromIEW->nextMicroPC[tid]; + toIEW->commitInfo[tid].pc = fromIEW->pc[tid]; toIEW->commitInfo[tid].mispredPC = fromIEW->mispredPC[tid]; @@ -880,10 +872,7 @@ DefaultCommit<Impl>::commitInsts() // Record that the number of ROB entries has changed. changedROBNumEntries[tid] = true; } else { - PC[tid] = head_inst->readPC(); - nextPC[tid] = head_inst->readNextPC(); - nextNPC[tid] = head_inst->readNextNPC(); - nextMicroPC[tid] = head_inst->readNextMicroPC(); + pc[tid] = head_inst->pcState(); // Increment the total number of non-speculative instructions // executed. @@ -911,11 +900,7 @@ DefaultCommit<Impl>::commitInsts() cpu->instDone(tid); } - PC[tid] = nextPC[tid]; - nextPC[tid] = nextNPC[tid]; - nextNPC[tid] = nextNPC[tid] + sizeof(TheISA::MachInst); - microPC[tid] = nextMicroPC[tid]; - nextMicroPC[tid] = microPC[tid] + 1; + TheISA::advancePC(pc[tid], head_inst->staticInst); int count = 0; Addr oldpc; @@ -923,19 +908,19 @@ DefaultCommit<Impl>::commitInsts() // currently updating state while handling PC events. assert(!thread[tid]->inSyscall && !thread[tid]->trapPending); do { - oldpc = PC[tid]; + oldpc = pc[tid].instAddr(); cpu->system->pcEventQueue.service(thread[tid]->getTC()); count++; - } while (oldpc != PC[tid]); + } while (oldpc != pc[tid].instAddr()); if (count > 1) { DPRINTF(Commit, "PC skip function event, stopping commit\n"); break; } } else { - DPRINTF(Commit, "Unable to commit head instruction PC:%#x " + DPRINTF(Commit, "Unable to commit head instruction PC:%s " "[tid:%i] [sn:%i].\n", - head_inst->readPC(), tid ,head_inst->seqNum); + head_inst->pcState(), tid ,head_inst->seqNum); break; } } @@ -970,8 +955,8 @@ DefaultCommit<Impl>::commitHead(DynInstPtr &head_inst, unsigned inst_num) head_inst->isWriteBarrier()) { DPRINTF(Commit, "Encountered a barrier or non-speculative " - "instruction [sn:%lli] at the head of the ROB, PC %#x.\n", - head_inst->seqNum, head_inst->readPC()); + "instruction [sn:%lli] at the head of the ROB, PC %s.\n", + head_inst->seqNum, head_inst->pcState()); if (inst_num > 0 || iewStage->hasStoresToWB(tid)) { DPRINTF(Commit, "Waiting for all stores to writeback.\n"); @@ -994,8 +979,8 @@ DefaultCommit<Impl>::commitHead(DynInstPtr &head_inst, unsigned inst_num) } assert(head_inst->uncacheable()); - DPRINTF(Commit, "[sn:%lli]: Uncached load, PC %#x.\n", - head_inst->seqNum, head_inst->readPC()); + DPRINTF(Commit, "[sn:%lli]: Uncached load, PC %s.\n", + head_inst->seqNum, head_inst->pcState()); // Send back the non-speculative instruction's sequence // number. Tell the lsq to re-execute the load. @@ -1034,8 +1019,8 @@ DefaultCommit<Impl>::commitHead(DynInstPtr &head_inst, unsigned inst_num) #endif if (inst_fault != NoFault) { - DPRINTF(Commit, "Inst [sn:%lli] PC %#x has a fault\n", - head_inst->seqNum, head_inst->readPC()); + DPRINTF(Commit, "Inst [sn:%lli] PC %s has a fault\n", + head_inst->seqNum, head_inst->pcState()); if (iewStage->hasStoresToWB(tid) || inst_num > 0) { DPRINTF(Commit, "Stores outstanding, fault must wait.\n"); @@ -1081,7 +1066,6 @@ DefaultCommit<Impl>::commitHead(DynInstPtr &head_inst, unsigned inst_num) // Generate trap squash event. generateTrapEvent(tid); -// warn("%lli fault (%d) handled @ PC %08p", curTick, inst_fault->name(), head_inst->readPC()); return false; } @@ -1089,9 +1073,7 @@ DefaultCommit<Impl>::commitHead(DynInstPtr &head_inst, unsigned inst_num) #if FULL_SYSTEM if (thread[tid]->profile) { -// bool usermode = TheISA::inUserMode(thread[tid]->getTC()); -// thread[tid]->profilePC = usermode ? 1 : head_inst->readPC(); - thread[tid]->profilePC = head_inst->readPC(); + thread[tid]->profilePC = head_inst->instAddr(); ProfileNode *node = thread[tid]->profile->consume(thread[tid]->getTC(), head_inst->staticInst); @@ -1101,7 +1083,7 @@ DefaultCommit<Impl>::commitHead(DynInstPtr &head_inst, unsigned inst_num) if (CPA::available()) { if (head_inst->isControl()) { ThreadContext *tc = thread[tid]->getTC(); - CPA::cpa()->swAutoBegin(tc, head_inst->readNextPC()); + CPA::cpa()->swAutoBegin(tc, head_inst->nextInstAddr()); } } #endif @@ -1154,8 +1136,8 @@ DefaultCommit<Impl>::getInsts() commitStatus[tid] != TrapPending) { changedROBNumEntries[tid] = true; - DPRINTF(Commit, "Inserting PC %#x [sn:%i] [tid:%i] into ROB.\n", - inst->readPC(), inst->seqNum, tid); + DPRINTF(Commit, "Inserting PC %s [sn:%i] [tid:%i] into ROB.\n", + inst->pcState(), inst->seqNum, tid); rob->insertInst(inst); @@ -1163,9 +1145,9 @@ DefaultCommit<Impl>::getInsts() youngestSeqNum[tid] = inst->seqNum; } else { - DPRINTF(Commit, "Instruction PC %#x [sn:%i] [tid:%i] was " + DPRINTF(Commit, "Instruction PC %s [sn:%i] [tid:%i] was " "squashed, skipping.\n", - inst->readPC(), inst->seqNum, tid); + inst->pcState(), inst->seqNum, tid); } } } @@ -1181,14 +1163,14 @@ DefaultCommit<Impl>::skidInsert() DynInstPtr inst = fromRename->insts[inst_num]; if (!inst->isSquashed()) { - DPRINTF(Commit, "Inserting PC %#x [sn:%i] [tid:%i] into ", - "skidBuffer.\n", inst->readPC(), inst->seqNum, + DPRINTF(Commit, "Inserting PC %s [sn:%i] [tid:%i] into ", + "skidBuffer.\n", inst->pcState(), inst->seqNum, inst->threadNumber); skidBuffer.push(inst); } else { - DPRINTF(Commit, "Instruction PC %#x [sn:%i] [tid:%i] was " + DPRINTF(Commit, "Instruction PC %s [sn:%i] [tid:%i] was " "squashed, skipping.\n", - inst->readPC(), inst->seqNum, inst->threadNumber); + inst->pcState(), inst->seqNum, inst->threadNumber); } } } @@ -1204,10 +1186,10 @@ DefaultCommit<Impl>::markCompletedInsts() ++inst_num) { if (!fromIEW->insts[inst_num]->isSquashed()) { - DPRINTF(Commit, "[tid:%i]: Marking PC %#x, [sn:%lli] ready " + DPRINTF(Commit, "[tid:%i]: Marking PC %s, [sn:%lli] ready " "within ROB.\n", fromIEW->insts[inst_num]->threadNumber, - fromIEW->insts[inst_num]->readPC(), + fromIEW->insts[inst_num]->pcState(), fromIEW->insts[inst_num]->seqNum); // Mark the instruction as ready to commit. diff --git a/src/cpu/o3/cpu.cc b/src/cpu/o3/cpu.cc index 8e9f3ef5d..21c5cc706 100644 --- a/src/cpu/o3/cpu.cc +++ b/src/cpu/o3/cpu.cc @@ -732,9 +732,7 @@ FullO3CPU<Impl>::insertThread(ThreadID tid) //this->copyFromTC(tid); //Set PC/NPC/NNPC - setPC(src_tc->readPC(), tid); - setNextPC(src_tc->readNextPC(), tid); - setNextNPC(src_tc->readNextNPC(), tid); + pcState(src_tc->pcState(), tid); src_tc->setStatus(ThreadContext::Active); @@ -778,7 +776,7 @@ FullO3CPU<Impl>::removeThread(ThreadID tid) // Squash Throughout Pipeline InstSeqNum squash_seq_num = commit.rob->readHeadInst(tid)->seqNum; - fetch.squash(0, sizeof(TheISA::MachInst), 0, squash_seq_num, tid); + fetch.squash(0, squash_seq_num, tid); decode.squash(tid); rename.squash(squash_seq_num, tid); iew.squash(tid); @@ -1306,73 +1304,38 @@ FullO3CPU<Impl>::setArchFloatRegInt(int reg_idx, uint64_t val, ThreadID tid) } template <class Impl> -uint64_t -FullO3CPU<Impl>::readPC(ThreadID tid) +TheISA::PCState +FullO3CPU<Impl>::pcState(ThreadID tid) { - return commit.readPC(tid); + return commit.pcState(tid); } template <class Impl> void -FullO3CPU<Impl>::setPC(Addr new_PC, ThreadID tid) -{ - commit.setPC(new_PC, tid); -} - -template <class Impl> -uint64_t -FullO3CPU<Impl>::readMicroPC(ThreadID tid) +FullO3CPU<Impl>::pcState(const TheISA::PCState &val, ThreadID tid) { - return commit.readMicroPC(tid); + commit.pcState(val, tid); } template <class Impl> -void -FullO3CPU<Impl>::setMicroPC(Addr new_PC, ThreadID tid) +Addr +FullO3CPU<Impl>::instAddr(ThreadID tid) { - commit.setMicroPC(new_PC, tid); + return commit.instAddr(tid); } template <class Impl> -uint64_t -FullO3CPU<Impl>::readNextPC(ThreadID tid) +Addr +FullO3CPU<Impl>::nextInstAddr(ThreadID tid) { - return commit.readNextPC(tid); + return commit.nextInstAddr(tid); } template <class Impl> -void -FullO3CPU<Impl>::setNextPC(uint64_t val, ThreadID tid) -{ - commit.setNextPC(val, tid); -} - -template <class Impl> -uint64_t -FullO3CPU<Impl>::readNextNPC(ThreadID tid) -{ - return commit.readNextNPC(tid); -} - -template <class Impl> -void -FullO3CPU<Impl>::setNextNPC(uint64_t val, ThreadID tid) -{ - commit.setNextNPC(val, tid); -} - -template <class Impl> -uint64_t -FullO3CPU<Impl>::readNextMicroPC(ThreadID tid) -{ - return commit.readNextMicroPC(tid); -} - -template <class Impl> -void -FullO3CPU<Impl>::setNextMicroPC(Addr new_PC, ThreadID tid) +MicroPC +FullO3CPU<Impl>::microPC(ThreadID tid) { - commit.setNextMicroPC(new_PC, tid); + return commit.microPC(tid); } template <class Impl> @@ -1419,9 +1382,9 @@ template <class Impl> void FullO3CPU<Impl>::removeFrontInst(DynInstPtr &inst) { - DPRINTF(O3CPU, "Removing committed instruction [tid:%i] PC %#x " + DPRINTF(O3CPU, "Removing committed instruction [tid:%i] PC %s " "[sn:%lli]\n", - inst->threadNumber, inst->readPC(), inst->seqNum); + inst->threadNumber, inst->pcState(), inst->seqNum); removeInstsThisCycle = true; @@ -1509,10 +1472,10 @@ FullO3CPU<Impl>::squashInstIt(const ListIt &instIt, ThreadID tid) { if ((*instIt)->threadNumber == tid) { DPRINTF(O3CPU, "Squashing instruction, " - "[tid:%i] [sn:%lli] PC %#x\n", + "[tid:%i] [sn:%lli] PC %s\n", (*instIt)->threadNumber, (*instIt)->seqNum, - (*instIt)->readPC()); + (*instIt)->pcState()); // Mark it as squashed. (*instIt)->setSquashed(); @@ -1530,10 +1493,10 @@ FullO3CPU<Impl>::cleanUpRemovedInsts() { while (!removeList.empty()) { DPRINTF(O3CPU, "Removing instruction, " - "[tid:%i] [sn:%lli] PC %#x\n", + "[tid:%i] [sn:%lli] PC %s\n", (*removeList.front())->threadNumber, (*removeList.front())->seqNum, - (*removeList.front())->readPC()); + (*removeList.front())->pcState()); instList.erase(removeList.front()); @@ -1563,7 +1526,7 @@ FullO3CPU<Impl>::dumpInsts() while (inst_list_it != instList.end()) { cprintf("Instruction:%i\nPC:%#x\n[tid:%i]\n[sn:%lli]\nIssued:%i\n" "Squashed:%i\n\n", - num, (*inst_list_it)->readPC(), (*inst_list_it)->threadNumber, + num, (*inst_list_it)->instAddr(), (*inst_list_it)->threadNumber, (*inst_list_it)->seqNum, (*inst_list_it)->isIssued(), (*inst_list_it)->isSquashed()); inst_list_it++; diff --git a/src/cpu/o3/cpu.hh b/src/cpu/o3/cpu.hh index 57c07a9ec..2669016ff 100644 --- a/src/cpu/o3/cpu.hh +++ b/src/cpu/o3/cpu.hh @@ -444,35 +444,20 @@ class FullO3CPU : public BaseO3CPU void setArchFloatRegInt(int reg_idx, uint64_t val, ThreadID tid); - /** Reads the commit PC of a specific thread. */ - Addr readPC(ThreadID tid); + /** Sets the commit PC state of a specific thread. */ + void pcState(const TheISA::PCState &newPCState, ThreadID tid); - /** Sets the commit PC of a specific thread. */ - void setPC(Addr new_PC, ThreadID tid); + /** Reads the commit PC state of a specific thread. */ + TheISA::PCState pcState(ThreadID tid); - /** Reads the commit micro PC of a specific thread. */ - Addr readMicroPC(ThreadID tid); + /** Reads the commit PC of a specific thread. */ + Addr instAddr(ThreadID tid); - /** Sets the commmit micro PC of a specific thread. */ - void setMicroPC(Addr new_microPC, ThreadID tid); + /** Reads the commit micro PC of a specific thread. */ + MicroPC microPC(ThreadID tid); /** Reads the next PC of a specific thread. */ - Addr readNextPC(ThreadID tid); - - /** Sets the next PC of a specific thread. */ - void setNextPC(Addr val, ThreadID tid); - - /** Reads the next NPC of a specific thread. */ - Addr readNextNPC(ThreadID tid); - - /** Sets the next NPC of a specific thread. */ - void setNextNPC(Addr val, ThreadID tid); - - /** Reads the commit next micro PC of a specific thread. */ - Addr readNextMicroPC(ThreadID tid); - - /** Sets the commit next micro PC of a specific thread. */ - void setNextMicroPC(Addr val, ThreadID tid); + Addr nextInstAddr(ThreadID tid); /** Initiates a squash of all in-flight instructions for a given * thread. The source of the squash is an external update of diff --git a/src/cpu/o3/decode_impl.hh b/src/cpu/o3/decode_impl.hh index 1b76de132..5063e0f07 100644 --- a/src/cpu/o3/decode_impl.hh +++ b/src/cpu/o3/decode_impl.hh @@ -264,29 +264,16 @@ template<class Impl> void DefaultDecode<Impl>::squash(DynInstPtr &inst, ThreadID tid) { - DPRINTF(Decode, "[tid:%i]: [sn:%i] Squashing due to incorrect branch prediction " - "detected at decode.\n", tid, inst->seqNum); + DPRINTF(Decode, "[tid:%i]: [sn:%i] Squashing due to incorrect branch " + "prediction detected at decode.\n", tid, inst->seqNum); // Send back mispredict information. toFetch->decodeInfo[tid].branchMispredict = true; toFetch->decodeInfo[tid].predIncorrect = true; toFetch->decodeInfo[tid].squash = true; toFetch->decodeInfo[tid].doneSeqNum = inst->seqNum; - toFetch->decodeInfo[tid].nextMicroPC = inst->readMicroPC(); - -#if ISA_HAS_DELAY_SLOT - toFetch->decodeInfo[tid].nextPC = inst->readPC() + sizeof(TheISA::MachInst); - toFetch->decodeInfo[tid].nextNPC = inst->branchTarget(); - toFetch->decodeInfo[tid].branchTaken = inst->readNextNPC() != - (inst->readNextPC() + sizeof(TheISA::MachInst)); -#else toFetch->decodeInfo[tid].nextPC = inst->branchTarget(); - toFetch->decodeInfo[tid].nextNPC = - inst->branchTarget() + sizeof(TheISA::MachInst); - toFetch->decodeInfo[tid].branchTaken = - inst->readNextPC() != (inst->readPC() + sizeof(TheISA::MachInst)); -#endif - + toFetch->decodeInfo[tid].branchTaken = inst->pcState().branching(); InstSeqNum squash_seq_num = inst->seqNum; @@ -382,8 +369,8 @@ DefaultDecode<Impl>::skidInsert(ThreadID tid) assert(tid == inst->threadNumber); - DPRINTF(Decode,"Inserting [sn:%lli] PC:%#x into decode skidBuffer %i\n", - inst->seqNum, inst->readPC(), inst->threadNumber); + DPRINTF(Decode,"Inserting [sn:%lli] PC: %s into decode skidBuffer %i\n", + inst->seqNum, inst->pcState(), inst->threadNumber); skidBuffer[tid].push(inst); } @@ -681,13 +668,12 @@ DefaultDecode<Impl>::decodeInsts(ThreadID tid) insts_to_decode.pop(); DPRINTF(Decode, "[tid:%u]: Processing instruction [sn:%lli] with " - "PC %#x\n", - tid, inst->seqNum, inst->readPC()); + "PC %s\n", tid, inst->seqNum, inst->pcState()); if (inst->isSquashed()) { - DPRINTF(Decode, "[tid:%u]: Instruction %i with PC %#x is " + DPRINTF(Decode, "[tid:%u]: Instruction %i with PC %s is " "squashed, skipping.\n", - tid, inst->seqNum, inst->readPC()); + tid, inst->seqNum, inst->pcState()); ++decodeSquashedInsts; @@ -717,9 +703,6 @@ DefaultDecode<Impl>::decodeInsts(ThreadID tid) // Ensure that if it was predicted as a branch, it really is a // branch. if (inst->readPredTaken() && !inst->isControl()) { - DPRINTF(Decode, "PredPC : %#x != NextPC: %#x\n", - inst->readPredPC(), inst->readNextPC() + 4); - panic("Instruction predicted as a branch!"); ++decodeControlMispred; @@ -735,26 +718,18 @@ DefaultDecode<Impl>::decodeInsts(ThreadID tid) if (inst->isDirectCtrl() && inst->isUncondCtrl()) { ++decodeBranchResolved; - if (inst->branchTarget() != inst->readPredPC()) { + if (!(inst->branchTarget() == inst->readPredTarg())) { ++decodeBranchMispred; // Might want to set some sort of boolean and just do // a check at the end squash(inst, inst->threadNumber); - Addr target = inst->branchTarget(); + TheISA::PCState target = inst->branchTarget(); -#if ISA_HAS_DELAY_SLOT - DPRINTF(Decode, "[sn:%i]: Updating predictions: PredPC: %#x PredNextPC: %#x\n", - inst->seqNum, inst->readPC() + sizeof(TheISA::MachInst), target); - - //The micro pc after an instruction level branch should be 0 - inst->setPredTarg(inst->readPC() + sizeof(TheISA::MachInst), target, 0); -#else - DPRINTF(Decode, "[sn:%i]: Updating predictions: PredPC: %#x PredNextPC: %#x\n", - inst->seqNum, target, target + sizeof(TheISA::MachInst)); + DPRINTF(Decode, "[sn:%i]: Updating predictions: PredPC: %s\n", + inst->seqNum, target); //The micro pc after an instruction level branch should be 0 - inst->setPredTarg(target, target + sizeof(TheISA::MachInst), 0); -#endif + inst->setPredTarg(target); break; } } diff --git a/src/cpu/o3/dep_graph.hh b/src/cpu/o3/dep_graph.hh index c19fd0abf..804b3f9cd 100644 --- a/src/cpu/o3/dep_graph.hh +++ b/src/cpu/o3/dep_graph.hh @@ -251,8 +251,8 @@ DependencyGraph<DynInstPtr>::dump() curr = &dependGraph[i]; if (curr->inst) { - cprintf("dependGraph[%i]: producer: %#x [sn:%lli] consumer: ", - i, curr->inst->readPC(), curr->inst->seqNum); + cprintf("dependGraph[%i]: producer: %s [sn:%lli] consumer: ", + i, curr->inst->pcState(), curr->inst->seqNum); } else { cprintf("dependGraph[%i]: No producer. consumer: ", i); } @@ -260,8 +260,8 @@ DependencyGraph<DynInstPtr>::dump() while (curr->next != NULL) { curr = curr->next; - cprintf("%#x [sn:%lli] ", - curr->inst->readPC(), curr->inst->seqNum); + cprintf("%s [sn:%lli] ", + curr->inst->pcState(), curr->inst->seqNum); } cprintf("\n"); diff --git a/src/cpu/o3/dyn_inst.hh b/src/cpu/o3/dyn_inst.hh index e1279f82b..6bd1f9fad 100644 --- a/src/cpu/o3/dyn_inst.hh +++ b/src/cpu/o3/dyn_inst.hh @@ -74,14 +74,14 @@ class BaseO3DynInst : public BaseDynInst<Impl> public: /** BaseDynInst constructor given a binary instruction. */ - BaseO3DynInst(StaticInstPtr staticInst, Addr PC, Addr NPC, Addr microPC, - Addr Pred_PC, Addr Pred_NPC, Addr Pred_MicroPC, - InstSeqNum seq_num, O3CPU *cpu); + BaseO3DynInst(StaticInstPtr staticInst, + TheISA::PCState pc, TheISA::PCState predPC, + InstSeqNum seq_num, O3CPU *cpu); /** BaseDynInst constructor given a binary instruction. */ - BaseO3DynInst(ExtMachInst inst, Addr PC, Addr NPC, Addr microPC, - Addr Pred_PC, Addr Pred_NPC, Addr Pred_MicroPC, - InstSeqNum seq_num, O3CPU *cpu); + BaseO3DynInst(ExtMachInst inst, + TheISA::PCState pc, TheISA::PCState predPC, + InstSeqNum seq_num, O3CPU *cpu); /** BaseDynInst constructor given a static inst pointer. */ BaseO3DynInst(StaticInstPtr &_staticInst); diff --git a/src/cpu/o3/dyn_inst_impl.hh b/src/cpu/o3/dyn_inst_impl.hh index 3e015d962..6f7c23b2f 100644 --- a/src/cpu/o3/dyn_inst_impl.hh +++ b/src/cpu/o3/dyn_inst_impl.hh @@ -33,24 +33,18 @@ template <class Impl> BaseO3DynInst<Impl>::BaseO3DynInst(StaticInstPtr staticInst, - Addr PC, Addr NPC, Addr microPC, - Addr Pred_PC, Addr Pred_NPC, - Addr Pred_MicroPC, + TheISA::PCState pc, TheISA::PCState predPC, InstSeqNum seq_num, O3CPU *cpu) - : BaseDynInst<Impl>(staticInst, PC, NPC, microPC, - Pred_PC, Pred_NPC, Pred_MicroPC, seq_num, cpu) + : BaseDynInst<Impl>(staticInst, pc, predPC, seq_num, cpu) { initVars(); } template <class Impl> BaseO3DynInst<Impl>::BaseO3DynInst(ExtMachInst inst, - Addr PC, Addr NPC, Addr microPC, - Addr Pred_PC, Addr Pred_NPC, - Addr Pred_MicroPC, + TheISA::PCState pc, TheISA::PCState predPC, InstSeqNum seq_num, O3CPU *cpu) - : BaseDynInst<Impl>(inst, PC, NPC, microPC, - Pred_PC, Pred_NPC, Pred_MicroPC, seq_num, cpu) + : BaseDynInst<Impl>(inst, pc, predPC, seq_num, cpu) { initVars(); } @@ -131,15 +125,17 @@ BaseO3DynInst<Impl>::hwrei() { #if THE_ISA == ALPHA_ISA // Can only do a hwrei when in pal mode. - if (!(this->readPC() & 0x3)) + if (!(this->instAddr() & 0x3)) return new AlphaISA::UnimplementedOpcodeFault; // Set the next PC based on the value of the EXC_ADDR IPR. - this->setNextPC(this->cpu->readMiscRegNoEffect(AlphaISA::IPR_EXC_ADDR, - this->threadNumber)); + AlphaISA::PCState pc = this->pcState(); + pc.npc(this->cpu->readMiscRegNoEffect(AlphaISA::IPR_EXC_ADDR, + this->threadNumber)); + this->pcState(pc); if (CPA::available()) { ThreadContext *tc = this->cpu->tcBase(this->threadNumber); - CPA::cpa()->swAutoBegin(tc, this->readNextPC()); + CPA::cpa()->swAutoBegin(tc, this->nextInstAddr()); } // Tell CPU to clear any state it needs to if a hwrei is taken. @@ -175,11 +171,11 @@ BaseO3DynInst<Impl>::syscall(int64_t callnum) // HACK: check CPU's nextPC before and after syscall. If it // changes, update this instruction's nextPC because the syscall // must have changed the nextPC. - Addr cpu_next_pc = this->cpu->readNextPC(this->threadNumber); + TheISA::PCState curPC = this->cpu->pcState(this->threadNumber); this->cpu->syscall(callnum, this->threadNumber); - Addr new_next_pc = this->cpu->readNextPC(this->threadNumber); - if (cpu_next_pc != new_next_pc) { - this->setNextPC(new_next_pc); + TheISA::PCState newPC = this->cpu->pcState(this->threadNumber); + if (!(curPC == newPC)) { + this->pcState(newPC); } } #endif diff --git a/src/cpu/o3/fetch.hh b/src/cpu/o3/fetch.hh index d881c291f..22e9e51b4 100644 --- a/src/cpu/o3/fetch.hh +++ b/src/cpu/o3/fetch.hh @@ -229,7 +229,7 @@ class DefaultFetch * @param next_NPC Used for ISAs which use delay slots. * @return Whether or not a branch was predicted as taken. */ - bool lookupAndUpdateNextPC(DynInstPtr &inst, Addr &next_PC, Addr &next_NPC, Addr &next_MicroPC); + bool lookupAndUpdateNextPC(DynInstPtr &inst, TheISA::PCState &pc); /** * Fetches the cache line that contains fetch_PC. Returns any @@ -244,14 +244,12 @@ class DefaultFetch bool fetchCacheLine(Addr fetch_PC, Fault &ret_fault, ThreadID tid); /** Squashes a specific thread and resets the PC. */ - inline void doSquash(const Addr &new_PC, const Addr &new_NPC, - const Addr &new_MicroPC, ThreadID tid); + inline void doSquash(const TheISA::PCState &newPC, ThreadID tid); /** Squashes a specific thread and resets the PC. Also tells the CPU to * remove any instructions between fetch and decode that should be sqaushed. */ - void squashFromDecode(const Addr &new_PC, const Addr &new_NPC, - const Addr &new_MicroPC, + void squashFromDecode(const TheISA::PCState &newPC, const InstSeqNum &seq_num, ThreadID tid); /** Checks if a thread is stalled. */ @@ -266,8 +264,7 @@ class DefaultFetch * remove any instructions that are not in the ROB. The source of this * squash should be the commit stage. */ - void squash(const Addr &new_PC, const Addr &new_NPC, - const Addr &new_MicroPC, + void squash(const TheISA::PCState &newPC, const InstSeqNum &seq_num, ThreadID tid); /** Ticks the fetch stage, processing all inputs signals and fetching @@ -348,14 +345,7 @@ class DefaultFetch /** Predecoder. */ TheISA::Predecoder predecoder; - /** Per-thread fetch PC. */ - Addr PC[Impl::MaxThreads]; - - /** Per-thread fetch micro PC. */ - Addr microPC[Impl::MaxThreads]; - - /** Per-thread next PC. */ - Addr nextPC[Impl::MaxThreads]; + TheISA::PCState pc[Impl::MaxThreads]; /** Memory request used to access cache. */ RequestPtr memReq[Impl::MaxThreads]; diff --git a/src/cpu/o3/fetch_impl.hh b/src/cpu/o3/fetch_impl.hh index 3f8f84cab..bbd9ce4a2 100644 --- a/src/cpu/o3/fetch_impl.hh +++ b/src/cpu/o3/fetch_impl.hh @@ -316,9 +316,7 @@ DefaultFetch<Impl>::initStage() { // Setup PC and nextPC with initial state. for (ThreadID tid = 0; tid < numThreads; tid++) { - PC[tid] = cpu->readPC(tid); - nextPC[tid] = cpu->readNextPC(tid); - microPC[tid] = cpu->readMicroPC(tid); + pc[tid] = cpu->pcState(tid); } for (ThreadID tid = 0; tid < numThreads; tid++) { @@ -445,9 +443,7 @@ DefaultFetch<Impl>::takeOverFrom() stalls[i].rename = 0; stalls[i].iew = 0; stalls[i].commit = 0; - PC[i] = cpu->readPC(i); - nextPC[i] = cpu->readNextPC(i); - microPC[i] = cpu->readMicroPC(i); + pc[i] = cpu->pcState(i); fetchStatus[i] = Running; } numInst = 0; @@ -496,8 +492,8 @@ DefaultFetch<Impl>::switchToInactive() template <class Impl> bool -DefaultFetch<Impl>::lookupAndUpdateNextPC(DynInstPtr &inst, Addr &next_PC, - Addr &next_NPC, Addr &next_MicroPC) +DefaultFetch<Impl>::lookupAndUpdateNextPC( + DynInstPtr &inst, TheISA::PCState &nextPC) { // Do branch prediction check here. // A bit of a misnomer...next_PC is actually the current PC until @@ -505,51 +501,26 @@ DefaultFetch<Impl>::lookupAndUpdateNextPC(DynInstPtr &inst, Addr &next_PC, bool predict_taken; if (!inst->isControl()) { - if (inst->isMicroop() && !inst->isLastMicroop()) { - next_MicroPC++; - } else { - next_PC = next_NPC; - next_NPC = next_NPC + instSize; - next_MicroPC = 0; - } - inst->setPredTarg(next_PC, next_NPC, next_MicroPC); + TheISA::advancePC(nextPC, inst->staticInst); + inst->setPredTarg(nextPC); inst->setPredTaken(false); return false; } - //Assume for now that all control flow is to a different macroop which - //would reset the micro pc to 0. - next_MicroPC = 0; - ThreadID tid = inst->threadNumber; - Addr pred_PC = next_PC; - predict_taken = branchPred.predict(inst, pred_PC, tid); + predict_taken = branchPred.predict(inst, nextPC, tid); if (predict_taken) { - DPRINTF(Fetch, "[tid:%i]: [sn:%i]: Branch predicted to be taken to %#x.\n", - tid, inst->seqNum, pred_PC); + DPRINTF(Fetch, "[tid:%i]: [sn:%i]: Branch predicted to be taken to %s.\n", + tid, inst->seqNum, nextPC); } else { DPRINTF(Fetch, "[tid:%i]: [sn:%i]:Branch predicted to be not taken.\n", tid, inst->seqNum); } -#if ISA_HAS_DELAY_SLOT - next_PC = next_NPC; - if (predict_taken) - next_NPC = pred_PC; - else - next_NPC += instSize; -#else - if (predict_taken) - next_PC = pred_PC; - else - next_PC += instSize; - next_NPC = next_PC + instSize; -#endif - - DPRINTF(Fetch, "[tid:%i]: [sn:%i] Branch predicted to go to %#x and then %#x.\n", - tid, inst->seqNum, next_PC, next_NPC); - inst->setPredTarg(next_PC, next_NPC, next_MicroPC); + DPRINTF(Fetch, "[tid:%i]: [sn:%i] Branch predicted to go to %s.\n", + tid, inst->seqNum, nextPC); + inst->setPredTarg(nextPC); inst->setPredTaken(predict_taken); ++fetchedBranches; @@ -668,15 +639,12 @@ DefaultFetch<Impl>::fetchCacheLine(Addr fetch_PC, Fault &ret_fault, ThreadID tid template <class Impl> inline void -DefaultFetch<Impl>::doSquash(const Addr &new_PC, - const Addr &new_NPC, const Addr &new_microPC, ThreadID tid) +DefaultFetch<Impl>::doSquash(const TheISA::PCState &newPC, ThreadID tid) { - DPRINTF(Fetch, "[tid:%i]: Squashing, setting PC to: %#x, NPC to: %#x.\n", - tid, new_PC, new_NPC); + DPRINTF(Fetch, "[tid:%i]: Squashing, setting PC to: %s.\n", + tid, newPC); - PC[tid] = new_PC; - nextPC[tid] = new_NPC; - microPC[tid] = new_microPC; + pc[tid] = newPC; // Clear the icache miss if it's outstanding. if (fetchStatus[tid] == IcacheWaitResponse) { @@ -703,13 +671,12 @@ DefaultFetch<Impl>::doSquash(const Addr &new_PC, template<class Impl> void -DefaultFetch<Impl>::squashFromDecode(const Addr &new_PC, const Addr &new_NPC, - const Addr &new_MicroPC, +DefaultFetch<Impl>::squashFromDecode(const TheISA::PCState &newPC, const InstSeqNum &seq_num, ThreadID tid) { - DPRINTF(Fetch, "[tid:%i]: Squashing from decode.\n",tid); + DPRINTF(Fetch, "[tid:%i]: Squashing from decode.\n", tid); - doSquash(new_PC, new_NPC, new_MicroPC, tid); + doSquash(newPC, tid); // Tell the CPU to remove any instructions that are in flight between // fetch and decode. @@ -784,13 +751,12 @@ DefaultFetch<Impl>::updateFetchStatus() template <class Impl> void -DefaultFetch<Impl>::squash(const Addr &new_PC, const Addr &new_NPC, - const Addr &new_MicroPC, +DefaultFetch<Impl>::squash(const TheISA::PCState &newPC, const InstSeqNum &seq_num, ThreadID tid) { - DPRINTF(Fetch, "[tid:%u]: Squash from commit.\n",tid); + DPRINTF(Fetch, "[tid:%u]: Squash from commit.\n", tid); - doSquash(new_PC, new_NPC, new_MicroPC, tid); + doSquash(newPC, tid); // Tell the CPU to remove any instructions that are not in the ROB. cpu->removeInstsNotInROB(tid); @@ -903,16 +869,14 @@ DefaultFetch<Impl>::checkSignalsAndUpdate(ThreadID tid) DPRINTF(Fetch, "[tid:%u]: Squashing instructions due to squash " "from commit.\n",tid); // In any case, squash. - squash(fromCommit->commitInfo[tid].nextPC, - fromCommit->commitInfo[tid].nextNPC, - fromCommit->commitInfo[tid].nextMicroPC, + squash(fromCommit->commitInfo[tid].pc, fromCommit->commitInfo[tid].doneSeqNum, tid); // Also check if there's a mispredict that happened. if (fromCommit->commitInfo[tid].branchMispredict) { branchPred.squash(fromCommit->commitInfo[tid].doneSeqNum, - fromCommit->commitInfo[tid].nextPC, + fromCommit->commitInfo[tid].pc, fromCommit->commitInfo[tid].branchTaken, tid); } else { @@ -955,13 +919,10 @@ DefaultFetch<Impl>::checkSignalsAndUpdate(ThreadID tid) if (fetchStatus[tid] != Squashing) { - DPRINTF(Fetch, "Squashing from decode with PC = %#x, NPC = %#x\n", - fromDecode->decodeInfo[tid].nextPC, - fromDecode->decodeInfo[tid].nextNPC); + TheISA::PCState nextPC = fromDecode->decodeInfo[tid].nextPC; + DPRINTF(Fetch, "Squashing from decode with PC = %s\n", nextPC); // Squash unless we're already squashing squashFromDecode(fromDecode->decodeInfo[tid].nextPC, - fromDecode->decodeInfo[tid].nextNPC, - fromDecode->decodeInfo[tid].nextMicroPC, fromDecode->decodeInfo[tid].doneSeqNum, tid); @@ -1016,9 +977,7 @@ DefaultFetch<Impl>::fetch(bool &status_change) DPRINTF(Fetch, "Attempting to fetch from [tid:%i]\n", tid); // The current PC. - Addr fetch_PC = PC[tid]; - Addr fetch_NPC = nextPC[tid]; - Addr fetch_MicroPC = microPC[tid]; + TheISA::PCState fetchPC = pc[tid]; // Fault code for memory access. Fault fault = NoFault; @@ -1034,10 +993,9 @@ DefaultFetch<Impl>::fetch(bool &status_change) status_change = true; } else if (fetchStatus[tid] == Running) { DPRINTF(Fetch, "[tid:%i]: Attempting to translate and read " - "instruction, starting at PC %08p.\n", - tid, fetch_PC); + "instruction, starting at PC %s.\n", tid, fetchPC); - bool fetch_success = fetchCacheLine(fetch_PC, fault, tid); + bool fetch_success = fetchCacheLine(fetchPC.instAddr(), fault, tid); if (!fetch_success) { if (cacheBlocked) { ++icacheStallCycles; @@ -1075,20 +1033,21 @@ DefaultFetch<Impl>::fetch(bool &status_change) return; } - Addr next_PC = fetch_PC; - Addr next_NPC = fetch_NPC; - Addr next_MicroPC = fetch_MicroPC; + TheISA::PCState nextPC = fetchPC; InstSeqNum inst_seq; MachInst inst; ExtMachInst ext_inst; - // @todo: Fix this hack. - unsigned offset = (fetch_PC & cacheBlkMask) & ~3; StaticInstPtr staticInst = NULL; StaticInstPtr macroop = NULL; if (fault == NoFault) { + //XXX Masking out pal mode bit. This will break x86. Alpha needs + //to pull the pal mode bit ouf ot the instruction address. + unsigned offset = (fetchPC.instAddr() & ~1) - cacheDataPC[tid]; + assert(offset < cacheBlkSize); + // If the read of the first instruction was successful, then grab the // instructions from the rest of the cache line and put them into the // queue heading to decode. @@ -1104,15 +1063,6 @@ DefaultFetch<Impl>::fetch(bool &status_change) numInst < fetchWidth && !predicted_branch) { - // If we're branching after this instruction, quite fetching - // from the same block then. - predicted_branch = - (fetch_PC + sizeof(TheISA::MachInst) != fetch_NPC); - if (predicted_branch) { - DPRINTF(Fetch, "Branch detected with PC = %#x, NPC = %#x\n", - fetch_PC, fetch_NPC); - } - // Make sure this is a valid index. assert(offset <= cacheBlkSize - instSize); @@ -1122,16 +1072,16 @@ DefaultFetch<Impl>::fetch(bool &status_change) (&cacheData[tid][offset])); predecoder.setTC(cpu->thread[tid]->getTC()); - predecoder.moreBytes(fetch_PC, fetch_PC, inst); + predecoder.moreBytes(fetchPC, fetchPC.instAddr(), inst); - ext_inst = predecoder.getExtMachInst(); - staticInst = StaticInstPtr(ext_inst, fetch_PC); + ext_inst = predecoder.getExtMachInst(fetchPC); + staticInst = StaticInstPtr(ext_inst, fetchPC.instAddr()); if (staticInst->isMacroop()) macroop = staticInst; } do { if (macroop) { - staticInst = macroop->fetchMicroop(fetch_MicroPC); + staticInst = macroop->fetchMicroop(fetchPC.microPC()); if (staticInst->isLastMicroop()) macroop = NULL; } @@ -1141,8 +1091,7 @@ DefaultFetch<Impl>::fetch(bool &status_change) // Create a new DynInst from the instruction fetched. DynInstPtr instruction = new DynInst(staticInst, - fetch_PC, fetch_NPC, fetch_MicroPC, - next_PC, next_NPC, next_MicroPC, + fetchPC, nextPC, inst_seq, cpu); instruction->setTid(tid); @@ -1150,27 +1099,32 @@ DefaultFetch<Impl>::fetch(bool &status_change) instruction->setThreadState(cpu->thread[tid]); - DPRINTF(Fetch, "[tid:%i]: Instruction PC %#x (%d) created " - "[sn:%lli]\n", tid, instruction->readPC(), - instruction->readMicroPC(), inst_seq); + DPRINTF(Fetch, "[tid:%i]: Instruction PC %s (%d) created " + "[sn:%lli]\n", tid, instruction->pcState(), + instruction->microPC(), inst_seq); //DPRINTF(Fetch, "[tid:%i]: MachInst is %#x\n", tid, ext_inst); - DPRINTF(Fetch, "[tid:%i]: Instruction is: %s\n", - tid, instruction->staticInst->disassemble(fetch_PC)); + DPRINTF(Fetch, "[tid:%i]: Instruction is: %s\n", tid, + instruction->staticInst-> + disassemble(fetchPC.instAddr())); #if TRACING_ON instruction->traceData = cpu->getTracer()->getInstRecord(curTick, cpu->tcBase(tid), - instruction->staticInst, instruction->readPC(), - macroop, instruction->readMicroPC()); + instruction->staticInst, fetchPC, macroop); #else instruction->traceData = NULL; #endif - ///FIXME This needs to be more robust in dealing with delay slots + // If we're branching after this instruction, quite fetching + // from the same block then. + predicted_branch = fetchPC.branching(); predicted_branch |= - lookupAndUpdateNextPC(instruction, next_PC, next_NPC, next_MicroPC); + lookupAndUpdateNextPC(instruction, nextPC); + if (predicted_branch) { + DPRINTF(Fetch, "Branch detected with PC = %s\n", fetchPC); + } // Add instruction to the CPU's list of instructions. instruction->setInstListIt(cpu->addInst(instruction)); @@ -1185,9 +1139,7 @@ DefaultFetch<Impl>::fetch(bool &status_change) ++fetchedInsts; // Move to the next instruction, unless we have a branch. - fetch_PC = next_PC; - fetch_NPC = next_NPC; - fetch_MicroPC = next_MicroPC; + fetchPC = nextPC; if (instruction->isQuiesce()) { DPRINTF(Fetch, "Quiesce instruction encountered, halting fetch!", @@ -1202,7 +1154,8 @@ DefaultFetch<Impl>::fetch(bool &status_change) } while (staticInst->isMicroop() && !staticInst->isLastMicroop() && numInst < fetchWidth); - offset += instSize; + //XXX Masking out pal mode bit. + offset = (fetchPC.instAddr() & ~1) - cacheDataPC[tid]; } if (predicted_branch) { @@ -1224,10 +1177,8 @@ DefaultFetch<Impl>::fetch(bool &status_change) // Now that fetching is completed, update the PC to signify what the next // cycle will be. if (fault == NoFault) { - PC[tid] = next_PC; - nextPC[tid] = next_NPC; - microPC[tid] = next_MicroPC; - DPRINTF(Fetch, "[tid:%i]: Setting PC to %08p.\n", tid, next_PC); + pc[tid] = nextPC; + DPRINTF(Fetch, "[tid:%i]: Setting PC to %s.\n", tid, nextPC); } else { // We shouldn't be in an icache miss and also have a fault (an ITB // miss) @@ -1245,11 +1196,10 @@ DefaultFetch<Impl>::fetch(bool &status_change) ext_inst = TheISA::NoopMachInst; // Create a new DynInst from the dummy nop. - DynInstPtr instruction = new DynInst(ext_inst, - fetch_PC, fetch_NPC, fetch_MicroPC, - next_PC, next_NPC, next_MicroPC, + DynInstPtr instruction = new DynInst(ext_inst, fetchPC, nextPC, inst_seq, cpu); - instruction->setPredTarg(next_NPC, next_NPC + instSize, 0); + TheISA::advancePC(nextPC, instruction->staticInst); + instruction->setPredTarg(nextPC); instruction->setTid(tid); instruction->setASID(tid); @@ -1272,8 +1222,8 @@ DefaultFetch<Impl>::fetch(bool &status_change) fetchStatus[tid] = TrapPending; status_change = true; - DPRINTF(Fetch, "[tid:%i]: fault (%s) detected @ PC %08p", - tid, fault->name(), PC[tid]); + DPRINTF(Fetch, "[tid:%i]: fault (%s) detected @ PC %s", + tid, fault->name(), pc[tid]); } } diff --git a/src/cpu/o3/iew_impl.hh b/src/cpu/o3/iew_impl.hh index 2089913ef..608e70cde 100644 --- a/src/cpu/o3/iew_impl.hh +++ b/src/cpu/o3/iew_impl.hh @@ -449,27 +449,18 @@ template<class Impl> void DefaultIEW<Impl>::squashDueToBranch(DynInstPtr &inst, ThreadID tid) { - DPRINTF(IEW, "[tid:%i]: Squashing from a specific instruction, PC: %#x " - "[sn:%i].\n", tid, inst->readPC(), inst->seqNum); + DPRINTF(IEW, "[tid:%i]: Squashing from a specific instruction, PC: %s " + "[sn:%i].\n", tid, inst->pcState(), inst->seqNum); toCommit->squash[tid] = true; toCommit->squashedSeqNum[tid] = inst->seqNum; - toCommit->mispredPC[tid] = inst->readPC(); + toCommit->mispredPC[tid] = inst->instAddr(); toCommit->branchMispredict[tid] = true; -#if ISA_HAS_DELAY_SLOT - int instSize = sizeof(TheISA::MachInst); - toCommit->branchTaken[tid] = - !(inst->readNextPC() + instSize == inst->readNextNPC() && - (inst->readNextPC() == inst->readPC() + instSize || - inst->readNextPC() == inst->readPC() + 2 * instSize)); -#else - toCommit->branchTaken[tid] = inst->readNextPC() != - (inst->readPC() + sizeof(TheISA::MachInst)); -#endif - toCommit->nextPC[tid] = inst->readNextPC(); - toCommit->nextNPC[tid] = inst->readNextNPC(); - toCommit->nextMicroPC[tid] = inst->readNextMicroPC(); + toCommit->branchTaken[tid] = inst->pcState().branching(); + TheISA::PCState pc = inst->pcState(); + TheISA::advancePC(pc, inst->staticInst); + toCommit->pc[tid] = pc; toCommit->includeSquashInst[tid] = false; @@ -481,12 +472,13 @@ void DefaultIEW<Impl>::squashDueToMemOrder(DynInstPtr &inst, ThreadID tid) { DPRINTF(IEW, "[tid:%i]: Squashing from a specific instruction, " - "PC: %#x [sn:%i].\n", tid, inst->readPC(), inst->seqNum); + "PC: %s [sn:%i].\n", tid, inst->pcState(), inst->seqNum); toCommit->squash[tid] = true; toCommit->squashedSeqNum[tid] = inst->seqNum; - toCommit->nextPC[tid] = inst->readNextPC(); - toCommit->nextNPC[tid] = inst->readNextNPC(); + TheISA::PCState pc = inst->pcState(); + TheISA::advancePC(pc, inst->staticInst); + toCommit->pc[tid] = pc; toCommit->branchMispredict[tid] = false; toCommit->includeSquashInst[tid] = false; @@ -499,12 +491,11 @@ void DefaultIEW<Impl>::squashDueToMemBlocked(DynInstPtr &inst, ThreadID tid) { DPRINTF(IEW, "[tid:%i]: Memory blocked, squashing load and younger insts, " - "PC: %#x [sn:%i].\n", tid, inst->readPC(), inst->seqNum); + "PC: %s [sn:%i].\n", tid, inst->pcState(), inst->seqNum); toCommit->squash[tid] = true; toCommit->squashedSeqNum[tid] = inst->seqNum; - toCommit->nextPC[tid] = inst->readPC(); - toCommit->nextNPC[tid] = inst->readNextPC(); + toCommit->pc[tid] = inst->pcState(); toCommit->branchMispredict[tid] = false; // Must include the broadcasted SN in the squash. @@ -628,9 +619,9 @@ DefaultIEW<Impl>::skidInsert(ThreadID tid) insts[tid].pop(); - DPRINTF(Decode,"[tid:%i]: Inserting [sn:%lli] PC:%#x into " + DPRINTF(Decode,"[tid:%i]: Inserting [sn:%lli] PC:%s into " "dispatch skidBuffer %i\n",tid, inst->seqNum, - inst->readPC(),tid); + inst->pcState(),tid); skidBuffer[tid].push(inst); } @@ -986,9 +977,9 @@ DefaultIEW<Impl>::dispatchInsts(ThreadID tid) // Make sure there's a valid instruction there. assert(inst); - DPRINTF(IEW, "[tid:%i]: Issue: Adding PC %#x [sn:%lli] [tid:%i] to " + DPRINTF(IEW, "[tid:%i]: Issue: Adding PC %s [sn:%lli] [tid:%i] to " "IQ.\n", - tid, inst->readPC(), inst->seqNum, inst->threadNumber); + tid, inst->pcState(), inst->seqNum, inst->threadNumber); // Be sure to mark these instructions as ready so that the // commit stage can go ahead and execute them, and mark @@ -1165,7 +1156,7 @@ DefaultIEW<Impl>::printAvailableInsts() if (inst%3==0) std::cout << "\n\t"; - std::cout << "PC: " << fromIssue->insts[inst]->readPC() + std::cout << "PC: " << fromIssue->insts[inst]->pcState() << " TN: " << fromIssue->insts[inst]->threadNumber << " SN: " << fromIssue->insts[inst]->seqNum << " | "; @@ -1205,8 +1196,8 @@ DefaultIEW<Impl>::executeInsts() DynInstPtr inst = instQueue.getInstToExecute(); - DPRINTF(IEW, "Execute: Processing PC %#x, [tid:%i] [sn:%i].\n", - inst->readPC(), inst->threadNumber,inst->seqNum); + DPRINTF(IEW, "Execute: Processing PC %s, [tid:%i] [sn:%i].\n", + inst->pcState(), inst->threadNumber,inst->seqNum); // Check if the instruction is squashed; if so then skip it if (inst->isSquashed()) { @@ -1298,10 +1289,9 @@ DefaultIEW<Impl>::executeInsts() DPRINTF(IEW, "Execute: Branch mispredict detected.\n"); DPRINTF(IEW, "Predicted target was PC:%#x, NPC:%#x.\n", - inst->readPredPC(), inst->readPredNPC()); - DPRINTF(IEW, "Execute: Redirecting fetch to PC: %#x," - " NPC: %#x.\n", inst->readNextPC(), - inst->readNextNPC()); + inst->predInstAddr(), inst->predNextInstAddr()); + DPRINTF(IEW, "Execute: Redirecting fetch to PC: %s.\n", + inst->pcState(), inst->nextInstAddr()); // If incorrect, then signal the ROB that it must be squashed. squashDueToBranch(inst, tid); @@ -1318,16 +1308,11 @@ DefaultIEW<Impl>::executeInsts() DynInstPtr violator; violator = ldstQueue.getMemDepViolator(tid); - DPRINTF(IEW, "LDSTQ detected a violation. Violator PC: %#x " - "[sn:%lli], inst PC: %#x [sn:%lli]. Addr is: %#x.\n", - violator->readPC(), violator->seqNum, - inst->readPC(), inst->seqNum, inst->physEffAddr); - // Ensure the violating instruction is older than - // current squash -/* if (fetchRedirect[tid] && - violator->seqNum >= toCommit->squashedSeqNum[tid] + 1) - continue; -*/ + DPRINTF(IEW, "LDSTQ detected a violation. Violator PC: %s " + "[sn:%lli], inst PC: %s [sn:%lli]. Addr is: %#x.\n", + violator->pcState(), violator->seqNum, + inst->pcState(), inst->seqNum, inst->physEffAddr); + fetchRedirect[tid] = true; // Tell the instruction queue that a violation has occured. @@ -1342,8 +1327,8 @@ DefaultIEW<Impl>::executeInsts() fetchRedirect[tid] = true; DPRINTF(IEW, "Load operation couldn't execute because the " - "memory system is blocked. PC: %#x [sn:%lli]\n", - inst->readPC(), inst->seqNum); + "memory system is blocked. PC: %s [sn:%lli]\n", + inst->pcState(), inst->seqNum); squashDueToMemBlocked(inst, tid); } @@ -1356,8 +1341,9 @@ DefaultIEW<Impl>::executeInsts() DynInstPtr violator = ldstQueue.getMemDepViolator(tid); DPRINTF(IEW, "LDSTQ detected a violation. Violator PC: " - "%#x, inst PC: %#x. Addr is: %#x.\n", - violator->readPC(), inst->readPC(), inst->physEffAddr); + "%s, inst PC: %s. Addr is: %#x.\n", + violator->pcState(), inst->pcState(), + inst->physEffAddr); DPRINTF(IEW, "Violation will not be handled because " "already squashing\n"); @@ -1366,8 +1352,8 @@ DefaultIEW<Impl>::executeInsts() if (ldstQueue.loadBlocked(tid) && !ldstQueue.isLoadBlockedHandled(tid)) { DPRINTF(IEW, "Load operation couldn't execute because the " - "memory system is blocked. PC: %#x [sn:%lli]\n", - inst->readPC(), inst->seqNum); + "memory system is blocked. PC: %s [sn:%lli]\n", + inst->pcState(), inst->seqNum); DPRINTF(IEW, "Blocked load will not be handled because " "already squashing\n"); @@ -1408,8 +1394,8 @@ DefaultIEW<Impl>::writebackInsts() DynInstPtr inst = toCommit->insts[inst_num]; ThreadID tid = inst->threadNumber; - DPRINTF(IEW, "Sending instructions to commit, [sn:%lli] PC %#x.\n", - inst->seqNum, inst->readPC()); + DPRINTF(IEW, "Sending instructions to commit, [sn:%lli] PC %s.\n", + inst->seqNum, inst->pcState()); iewInstsToCommit[tid]++; @@ -1613,10 +1599,10 @@ DefaultIEW<Impl>::checkMisprediction(DynInstPtr &inst) DPRINTF(IEW, "Execute: Branch mispredict detected.\n"); DPRINTF(IEW, "Predicted target was PC:%#x, NPC:%#x.\n", - inst->readPredPC(), inst->readPredNPC()); + inst->predInstAddr(), inst->predNextInstAddr()); DPRINTF(IEW, "Execute: Redirecting fetch to PC: %#x," - " NPC: %#x.\n", inst->readNextPC(), - inst->readNextNPC()); + " NPC: %#x.\n", inst->nextInstAddr(), + inst->nextInstAddr()); // If incorrect, then signal the ROB that it must be squashed. squashDueToBranch(inst, tid); diff --git a/src/cpu/o3/inst_queue_impl.hh b/src/cpu/o3/inst_queue_impl.hh index 3d5eadf84..b944979f2 100644 --- a/src/cpu/o3/inst_queue_impl.hh +++ b/src/cpu/o3/inst_queue_impl.hh @@ -504,8 +504,8 @@ InstructionQueue<Impl>::insert(DynInstPtr &new_inst) // Make sure the instruction is valid assert(new_inst); - DPRINTF(IQ, "Adding instruction [sn:%lli] PC %#x to the IQ.\n", - new_inst->seqNum, new_inst->readPC()); + DPRINTF(IQ, "Adding instruction [sn:%lli] PC %s to the IQ.\n", + new_inst->seqNum, new_inst->pcState()); assert(freeEntries != 0); @@ -547,9 +547,9 @@ InstructionQueue<Impl>::insertNonSpec(DynInstPtr &new_inst) nonSpecInsts[new_inst->seqNum] = new_inst; - DPRINTF(IQ, "Adding non-speculative instruction [sn:%lli] PC %#x " + DPRINTF(IQ, "Adding non-speculative instruction [sn:%lli] PC %s " "to the IQ.\n", - new_inst->seqNum, new_inst->readPC()); + new_inst->seqNum, new_inst->pcState()); assert(freeEntries != 0); @@ -767,9 +767,9 @@ InstructionQueue<Impl>::scheduleReadyInsts() } } - DPRINTF(IQ, "Thread %i: Issuing instruction PC %#x " + DPRINTF(IQ, "Thread %i: Issuing instruction PC %s " "[sn:%lli]\n", - tid, issuing_inst->readPC(), + tid, issuing_inst->pcState(), issuing_inst->seqNum); readyInsts[op_class].pop(); @@ -910,7 +910,7 @@ InstructionQueue<Impl>::wakeDependents(DynInstPtr &completed_inst) while (dep_inst) { DPRINTF(IQ, "Waking up a dependent instruction, [sn:%lli] " - "PC%#x.\n", dep_inst->seqNum, dep_inst->readPC()); + "PC %s.\n", dep_inst->seqNum, dep_inst->pcState()); // Might want to give more information to the instruction // so that it knows which of its source registers is @@ -955,8 +955,8 @@ InstructionQueue<Impl>::addReadyMemInst(DynInstPtr &ready_inst) } DPRINTF(IQ, "Instruction is ready to issue, putting it onto " - "the ready list, PC %#x opclass:%i [sn:%lli].\n", - ready_inst->readPC(), op_class, ready_inst->seqNum); + "the ready list, PC %s opclass:%i [sn:%lli].\n", + ready_inst->pcState(), op_class, ready_inst->seqNum); } template <class Impl> @@ -981,8 +981,8 @@ InstructionQueue<Impl>::completeMemInst(DynInstPtr &completed_inst) { ThreadID tid = completed_inst->threadNumber; - DPRINTF(IQ, "Completing mem instruction PC:%#x [sn:%lli]\n", - completed_inst->readPC(), completed_inst->seqNum); + DPRINTF(IQ, "Completing mem instruction PC: %s [sn:%lli]\n", + completed_inst->pcState(), completed_inst->seqNum); ++freeEntries; @@ -1050,9 +1050,8 @@ InstructionQueue<Impl>::doSquash(ThreadID tid) (squashed_inst->isMemRef() && !squashed_inst->memOpDone)) { - DPRINTF(IQ, "[tid:%i]: Instruction [sn:%lli] PC %#x " - "squashed.\n", - tid, squashed_inst->seqNum, squashed_inst->readPC()); + DPRINTF(IQ, "[tid:%i]: Instruction [sn:%lli] PC %s squashed.\n", + tid, squashed_inst->seqNum, squashed_inst->pcState()); // Remove the instruction from the dependency list. if (!squashed_inst->isNonSpeculative() && @@ -1147,9 +1146,9 @@ InstructionQueue<Impl>::addToDependents(DynInstPtr &new_inst) if (src_reg >= numPhysRegs) { continue; } else if (regScoreboard[src_reg] == false) { - DPRINTF(IQ, "Instruction PC %#x has src reg %i that " + DPRINTF(IQ, "Instruction PC %s has src reg %i that " "is being added to the dependency chain.\n", - new_inst->readPC(), src_reg); + new_inst->pcState(), src_reg); dependGraph.insert(src_reg, new_inst); @@ -1157,9 +1156,9 @@ InstructionQueue<Impl>::addToDependents(DynInstPtr &new_inst) // was added to the dependency graph. return_val = true; } else { - DPRINTF(IQ, "Instruction PC %#x has src reg %i that " + DPRINTF(IQ, "Instruction PC %s has src reg %i that " "became ready before it reached the IQ.\n", - new_inst->readPC(), src_reg); + new_inst->pcState(), src_reg); // Mark a register ready within the instruction. new_inst->markSrcRegReady(src_reg_idx); } @@ -1228,8 +1227,8 @@ InstructionQueue<Impl>::addIfReady(DynInstPtr &inst) OpClass op_class = inst->opClass(); DPRINTF(IQ, "Instruction is ready to issue, putting it onto " - "the ready list, PC %#x opclass:%i [sn:%lli].\n", - inst->readPC(), op_class, inst->seqNum); + "the ready list, PC %s opclass:%i [sn:%lli].\n", + inst->pcState(), op_class, inst->seqNum); readyInsts[op_class].push(inst); @@ -1299,7 +1298,7 @@ InstructionQueue<Impl>::dumpLists() cprintf("Non speculative list: "); while (non_spec_it != non_spec_end_it) { - cprintf("%#x [sn:%lli]", (*non_spec_it).second->readPC(), + cprintf("%s [sn:%lli]", (*non_spec_it).second->pcState(), (*non_spec_it).second->seqNum); ++non_spec_it; } @@ -1348,9 +1347,9 @@ InstructionQueue<Impl>::dumpInsts() } } - cprintf("PC:%#x\n[sn:%lli]\n[tid:%i]\n" + cprintf("PC: %s\n[sn:%lli]\n[tid:%i]\n" "Issued:%i\nSquashed:%i\n", - (*inst_list_it)->readPC(), + (*inst_list_it)->pcState(), (*inst_list_it)->seqNum, (*inst_list_it)->threadNumber, (*inst_list_it)->isIssued(), @@ -1390,9 +1389,9 @@ InstructionQueue<Impl>::dumpInsts() } } - cprintf("PC:%#x\n[sn:%lli]\n[tid:%i]\n" + cprintf("PC: %s\n[sn:%lli]\n[tid:%i]\n" "Issued:%i\nSquashed:%i\n", - (*inst_list_it)->readPC(), + (*inst_list_it)->pcState(), (*inst_list_it)->seqNum, (*inst_list_it)->threadNumber, (*inst_list_it)->isIssued(), diff --git a/src/cpu/o3/lsq_unit.hh b/src/cpu/o3/lsq_unit.hh index a9047558d..372e76b71 100644 --- a/src/cpu/o3/lsq_unit.hh +++ b/src/cpu/o3/lsq_unit.hh @@ -530,8 +530,8 @@ LSQUnit<Impl>::read(Request *req, Request *sreqLow, Request *sreqHigh, (load_idx != loadHead || !load_inst->isAtCommit())) { iewStage->rescheduleMemInst(load_inst); ++lsqRescheduledLoads; - DPRINTF(LSQUnit, "Uncachable load [sn:%lli] PC %#x\n", - load_inst->seqNum, load_inst->readPC()); + DPRINTF(LSQUnit, "Uncachable load [sn:%lli] PC %s\n", + load_inst->seqNum, load_inst->pcState()); // Must delete request now that it wasn't handed off to // memory. This is quite ugly. @todo: Figure out the proper @@ -687,8 +687,8 @@ LSQUnit<Impl>::read(Request *req, Request *sreqLow, Request *sreqHigh, } // If there's no forwarding case, then go access memory - DPRINTF(LSQUnit, "Doing memory access for inst [sn:%lli] PC %#x\n", - load_inst->seqNum, load_inst->readPC()); + DPRINTF(LSQUnit, "Doing memory access for inst [sn:%lli] PC %s\n", + load_inst->seqNum, load_inst->pcState()); assert(!load_inst->memData); load_inst->memData = new uint8_t[64]; diff --git a/src/cpu/o3/lsq_unit_impl.hh b/src/cpu/o3/lsq_unit_impl.hh index 8aa7fe397..345d3ea69 100644 --- a/src/cpu/o3/lsq_unit_impl.hh +++ b/src/cpu/o3/lsq_unit_impl.hh @@ -352,8 +352,8 @@ LSQUnit<Impl>::insertLoad(DynInstPtr &load_inst) assert((loadTail + 1) % LQEntries != loadHead); assert(loads < LQEntries); - DPRINTF(LSQUnit, "Inserting load PC %#x, idx:%i [sn:%lli]\n", - load_inst->readPC(), loadTail, load_inst->seqNum); + DPRINTF(LSQUnit, "Inserting load PC %s, idx:%i [sn:%lli]\n", + load_inst->pcState(), loadTail, load_inst->seqNum); load_inst->lqIdx = loadTail; @@ -378,8 +378,8 @@ LSQUnit<Impl>::insertStore(DynInstPtr &store_inst) assert((storeTail + 1) % SQEntries != storeHead); assert(stores < SQEntries); - DPRINTF(LSQUnit, "Inserting store PC %#x, idx:%i [sn:%lli]\n", - store_inst->readPC(), storeTail, store_inst->seqNum); + DPRINTF(LSQUnit, "Inserting store PC %s, idx:%i [sn:%lli]\n", + store_inst->pcState(), storeTail, store_inst->seqNum); store_inst->sqIdx = storeTail; store_inst->lqIdx = loadTail; @@ -444,8 +444,8 @@ LSQUnit<Impl>::executeLoad(DynInstPtr &inst) // Execute a specific load. Fault load_fault = NoFault; - DPRINTF(LSQUnit, "Executing load PC %#x, [sn:%lli]\n", - inst->readPC(),inst->seqNum); + DPRINTF(LSQUnit, "Executing load PC %s, [sn:%lli]\n", + inst->pcState(),inst->seqNum); assert(!inst->isSquashed()); @@ -519,8 +519,8 @@ LSQUnit<Impl>::executeStore(DynInstPtr &store_inst) int store_idx = store_inst->sqIdx; - DPRINTF(LSQUnit, "Executing store PC %#x [sn:%lli]\n", - store_inst->readPC(), store_inst->seqNum); + DPRINTF(LSQUnit, "Executing store PC %s [sn:%lli]\n", + store_inst->pcState(), store_inst->seqNum); assert(!store_inst->isSquashed()); @@ -531,8 +531,8 @@ LSQUnit<Impl>::executeStore(DynInstPtr &store_inst) Fault store_fault = store_inst->initiateAcc(); if (storeQueue[store_idx].size == 0) { - DPRINTF(LSQUnit,"Fault on Store PC %#x, [sn:%lli],Size = 0\n", - store_inst->readPC(),store_inst->seqNum); + DPRINTF(LSQUnit,"Fault on Store PC %s, [sn:%lli],Size = 0\n", + store_inst->pcState(), store_inst->seqNum); return store_fault; } @@ -593,8 +593,8 @@ LSQUnit<Impl>::commitLoad() { assert(loadQueue[loadHead]); - DPRINTF(LSQUnit, "Committing head load instruction, PC %#x\n", - loadQueue[loadHead]->readPC()); + DPRINTF(LSQUnit, "Committing head load instruction, PC %s\n", + loadQueue[loadHead]->pcState()); loadQueue[loadHead] = NULL; @@ -631,8 +631,8 @@ LSQUnit<Impl>::commitStores(InstSeqNum &youngest_inst) break; } DPRINTF(LSQUnit, "Marking store as able to write back, PC " - "%#x [sn:%lli]\n", - storeQueue[store_idx].inst->readPC(), + "%s [sn:%lli]\n", + storeQueue[store_idx].inst->pcState(), storeQueue[store_idx].inst->seqNum); storeQueue[store_idx].canWB = true; @@ -757,9 +757,9 @@ LSQUnit<Impl>::writebackStores() req = sreqLow; } - DPRINTF(LSQUnit, "D-Cache: Writing back store idx:%i PC:%#x " + DPRINTF(LSQUnit, "D-Cache: Writing back store idx:%i PC:%s " "to Addr:%#x, data:%#x [sn:%lli]\n", - storeWBIdx, inst->readPC(), + storeWBIdx, inst->pcState(), req->getPaddr(), (int)*(inst->memData), inst->seqNum); @@ -861,9 +861,9 @@ LSQUnit<Impl>::squash(const InstSeqNum &squashed_num) decrLdIdx(load_idx); while (loads != 0 && loadQueue[load_idx]->seqNum > squashed_num) { - DPRINTF(LSQUnit,"Load Instruction PC %#x squashed, " + DPRINTF(LSQUnit,"Load Instruction PC %s squashed, " "[sn:%lli]\n", - loadQueue[load_idx]->readPC(), + loadQueue[load_idx]->pcState(), loadQueue[load_idx]->seqNum); if (isStalled() && load_idx == stallingLoadIdx) { @@ -906,9 +906,9 @@ LSQUnit<Impl>::squash(const InstSeqNum &squashed_num) break; } - DPRINTF(LSQUnit,"Store Instruction PC %#x squashed, " + DPRINTF(LSQUnit,"Store Instruction PC %s squashed, " "idx:%i [sn:%lli]\n", - storeQueue[store_idx].inst->readPC(), + storeQueue[store_idx].inst->pcState(), store_idx, storeQueue[store_idx].inst->seqNum); // I don't think this can happen. It should have been cleared @@ -1156,7 +1156,7 @@ LSQUnit<Impl>::dumpInsts() int load_idx = loadHead; while (load_idx != loadTail && loadQueue[load_idx]) { - cprintf("%#x ", loadQueue[load_idx]->readPC()); + cprintf("%s ", loadQueue[load_idx]->pcState()); incrLdIdx(load_idx); } @@ -1167,7 +1167,7 @@ LSQUnit<Impl>::dumpInsts() int store_idx = storeHead; while (store_idx != storeTail && storeQueue[store_idx].inst) { - cprintf("%#x ", storeQueue[store_idx].inst->readPC()); + cprintf("%s ", storeQueue[store_idx].inst->pcState()); incrStIdx(store_idx); } diff --git a/src/cpu/o3/mem_dep_unit_impl.hh b/src/cpu/o3/mem_dep_unit_impl.hh index 5f5e71624..fdea84ed5 100644 --- a/src/cpu/o3/mem_dep_unit_impl.hh +++ b/src/cpu/o3/mem_dep_unit_impl.hh @@ -171,7 +171,7 @@ MemDepUnit<MemDepPred, Impl>::insert(DynInstPtr &inst) storeBarrierSN); producing_store = storeBarrierSN; } else { - producing_store = depPred.checkInst(inst->readPC()); + producing_store = depPred.checkInst(inst->instAddr()); } MemDepEntryPtr store_entry = NULL; @@ -191,7 +191,7 @@ MemDepUnit<MemDepPred, Impl>::insert(DynInstPtr &inst) // are ready. if (!store_entry) { DPRINTF(MemDepUnit, "No dependency for inst PC " - "%#x [sn:%lli].\n", inst->readPC(), inst->seqNum); + "%s [sn:%lli].\n", inst->pcState(), inst->seqNum); inst_entry->memDepReady = true; @@ -203,8 +203,8 @@ MemDepUnit<MemDepPred, Impl>::insert(DynInstPtr &inst) } else { // Otherwise make the instruction dependent on the store/barrier. DPRINTF(MemDepUnit, "Adding to dependency list; " - "inst PC %#x is dependent on [sn:%lli].\n", - inst->readPC(), producing_store); + "inst PC %s is dependent on [sn:%lli].\n", + inst->pcState(), producing_store); if (inst->readyToIssue()) { inst_entry->regsReady = true; @@ -224,10 +224,10 @@ MemDepUnit<MemDepPred, Impl>::insert(DynInstPtr &inst) } if (inst->isStore()) { - DPRINTF(MemDepUnit, "Inserting store PC %#x [sn:%lli].\n", - inst->readPC(), inst->seqNum); + DPRINTF(MemDepUnit, "Inserting store PC %s [sn:%lli].\n", + inst->pcState(), inst->seqNum); - depPred.insertStore(inst->readPC(), inst->seqNum, inst->threadNumber); + depPred.insertStore(inst->instAddr(), inst->seqNum, inst->threadNumber); ++insertedStores; } else if (inst->isLoad()) { @@ -260,10 +260,10 @@ MemDepUnit<MemDepPred, Impl>::insertNonSpec(DynInstPtr &inst) // Might want to turn this part into an inline function or something. // It's shared between both insert functions. if (inst->isStore()) { - DPRINTF(MemDepUnit, "Inserting store PC %#x [sn:%lli].\n", - inst->readPC(), inst->seqNum); + DPRINTF(MemDepUnit, "Inserting store PC %s [sn:%lli].\n", + inst->pcState(), inst->seqNum); - depPred.insertStore(inst->readPC(), inst->seqNum, inst->threadNumber); + depPred.insertStore(inst->instAddr(), inst->seqNum, inst->threadNumber); ++insertedStores; } else if (inst->isLoad()) { @@ -313,8 +313,8 @@ void MemDepUnit<MemDepPred, Impl>::regsReady(DynInstPtr &inst) { DPRINTF(MemDepUnit, "Marking registers as ready for " - "instruction PC %#x [sn:%lli].\n", - inst->readPC(), inst->seqNum); + "instruction PC %s [sn:%lli].\n", + inst->pcState(), inst->seqNum); MemDepEntryPtr inst_entry = findInHash(inst); @@ -336,8 +336,8 @@ void MemDepUnit<MemDepPred, Impl>::nonSpecInstReady(DynInstPtr &inst) { DPRINTF(MemDepUnit, "Marking non speculative " - "instruction PC %#x as ready [sn:%lli].\n", - inst->readPC(), inst->seqNum); + "instruction PC %s as ready [sn:%lli].\n", + inst->pcState(), inst->seqNum); MemDepEntryPtr inst_entry = findInHash(inst); @@ -363,9 +363,8 @@ MemDepUnit<MemDepPred, Impl>::replay(DynInstPtr &inst) MemDepEntryPtr inst_entry = findInHash(temp_inst); - DPRINTF(MemDepUnit, "Replaying mem instruction PC %#x " - "[sn:%lli].\n", - temp_inst->readPC(), temp_inst->seqNum); + DPRINTF(MemDepUnit, "Replaying mem instruction PC %s [sn:%lli].\n", + temp_inst->pcState(), temp_inst->seqNum); moveToReady(inst_entry); @@ -377,9 +376,8 @@ template <class MemDepPred, class Impl> void MemDepUnit<MemDepPred, Impl>::completed(DynInstPtr &inst) { - DPRINTF(MemDepUnit, "Completed mem instruction PC %#x " - "[sn:%lli].\n", - inst->readPC(), inst->seqNum); + DPRINTF(MemDepUnit, "Completed mem instruction PC %s [sn:%lli].\n", + inst->pcState(), inst->seqNum); ThreadID tid = inst->threadNumber; @@ -507,10 +505,10 @@ MemDepUnit<MemDepPred, Impl>::violation(DynInstPtr &store_inst, DynInstPtr &violating_load) { DPRINTF(MemDepUnit, "Passing violating PCs to store sets," - " load: %#x, store: %#x\n", violating_load->readPC(), - store_inst->readPC()); + " load: %#x, store: %#x\n", violating_load->instAddr(), + store_inst->instAddr()); // Tell the memory dependence unit of the violation. - depPred.violation(violating_load->readPC(), store_inst->readPC()); + depPred.violation(violating_load->instAddr(), store_inst->instAddr()); } template <class MemDepPred, class Impl> @@ -518,9 +516,9 @@ void MemDepUnit<MemDepPred, Impl>::issue(DynInstPtr &inst) { DPRINTF(MemDepUnit, "Issuing instruction PC %#x [sn:%lli].\n", - inst->readPC(), inst->seqNum); + inst->instAddr(), inst->seqNum); - depPred.issued(inst->readPC(), inst->seqNum, inst->isStore()); + depPred.issued(inst->instAddr(), inst->seqNum, inst->isStore()); } template <class MemDepPred, class Impl> @@ -559,9 +557,9 @@ MemDepUnit<MemDepPred, Impl>::dumpLists() int num = 0; while (inst_list_it != instList[tid].end()) { - cprintf("Instruction:%i\nPC:%#x\n[sn:%i]\n[tid:%i]\nIssued:%i\n" + cprintf("Instruction:%i\nPC: %s\n[sn:%i]\n[tid:%i]\nIssued:%i\n" "Squashed:%i\n\n", - num, (*inst_list_it)->readPC(), + num, (*inst_list_it)->pcState(), (*inst_list_it)->seqNum, (*inst_list_it)->threadNumber, (*inst_list_it)->isIssued(), diff --git a/src/cpu/o3/rename_impl.hh b/src/cpu/o3/rename_impl.hh index 7f796c4c8..4b321d099 100644 --- a/src/cpu/o3/rename_impl.hh +++ b/src/cpu/o3/rename_impl.hh @@ -591,15 +591,14 @@ DefaultRename<Impl>::renameInsts(ThreadID tid) insts_to_rename.pop_front(); if (renameStatus[tid] == Unblocking) { - DPRINTF(Rename,"[tid:%u]: Removing [sn:%lli] PC:%#x from rename " - "skidBuffer\n", - tid, inst->seqNum, inst->readPC()); + DPRINTF(Rename,"[tid:%u]: Removing [sn:%lli] PC:%s from rename " + "skidBuffer\n", tid, inst->seqNum, inst->pcState()); } if (inst->isSquashed()) { - DPRINTF(Rename, "[tid:%u]: instruction %i with PC %#x is " - "squashed, skipping.\n", - tid, inst->seqNum, inst->readPC()); + DPRINTF(Rename, "[tid:%u]: instruction %i with PC %s is " + "squashed, skipping.\n", tid, inst->seqNum, + inst->pcState()); ++renameSquashedInsts; @@ -610,8 +609,7 @@ DefaultRename<Impl>::renameInsts(ThreadID tid) } DPRINTF(Rename, "[tid:%u]: Processing instruction [sn:%lli] with " - "PC %#x.\n", - tid, inst->seqNum, inst->readPC()); + "PC %s.\n", tid, inst->seqNum, inst->pcState()); // Handle serializeAfter/serializeBefore instructions. // serializeAfter marks the next instruction as serializeBefore. @@ -716,8 +714,8 @@ DefaultRename<Impl>::skidInsert(ThreadID tid) assert(tid == inst->threadNumber); - DPRINTF(Rename, "[tid:%u]: Inserting [sn:%lli] PC:%#x into Rename " - "skidBuffer\n", tid, inst->seqNum, inst->readPC()); + DPRINTF(Rename, "[tid:%u]: Inserting [sn:%lli] PC: %s into Rename " + "skidBuffer\n", tid, inst->seqNum, inst->pcState()); ++renameSkidInsts; @@ -731,7 +729,7 @@ DefaultRename<Impl>::skidInsert(ThreadID tid) for(it = skidBuffer[tid].begin(); it != skidBuffer[tid].end(); it++) { warn("[tid:%u]: %s [sn:%i].\n", tid, - (*it)->staticInst->disassemble(inst->readPC()), + (*it)->staticInst->disassemble(inst->instAddr()), (*it)->seqNum); } panic("Skidbuffer Exceeded Max Size"); @@ -1287,8 +1285,7 @@ DefaultRename<Impl>::checkSignalsAndUpdate(ThreadID tid) unblock(tid); DPRINTF(Rename, "[tid:%u]: Processing instruction [%lli] with " - "PC %#x.\n", - tid, serial_inst->seqNum, serial_inst->readPC()); + "PC %s.\n", tid, serial_inst->seqNum, serial_inst->pcState()); // Put instruction into queue here. serial_inst->clearSerializeBefore(); diff --git a/src/cpu/o3/rob_impl.hh b/src/cpu/o3/rob_impl.hh index cf0080b48..37f1c5504 100644 --- a/src/cpu/o3/rob_impl.hh +++ b/src/cpu/o3/rob_impl.hh @@ -204,7 +204,7 @@ ROB<Impl>::insertInst(DynInstPtr &inst) { assert(inst); - DPRINTF(ROB, "Adding inst PC %#x to the ROB.\n", inst->readPC()); + DPRINTF(ROB, "Adding inst PC %s to the ROB.\n", inst->pcState()); assert(numInstsInROB != numEntries); @@ -247,7 +247,7 @@ ROB<Impl>::retireHead(ThreadID tid) assert(head_inst->readyToCommit()); DPRINTF(ROB, "[tid:%u]: Retiring head instruction, " - "instruction PC %#x,[sn:%lli]\n", tid, head_inst->readPC(), + "instruction PC %s, [sn:%lli]\n", tid, head_inst->pcState(), head_inst->seqNum); --numInstsInROB; @@ -338,9 +338,9 @@ ROB<Impl>::doSquash(ThreadID tid) (*squashIt[tid])->seqNum > squashedSeqNum[tid]; ++numSquashed) { - DPRINTF(ROB, "[tid:%u]: Squashing instruction PC %#x, seq num %i.\n", + DPRINTF(ROB, "[tid:%u]: Squashing instruction PC %s, seq num %i.\n", (*squashIt[tid])->threadNumber, - (*squashIt[tid])->readPC(), + (*squashIt[tid])->pcState(), (*squashIt[tid])->seqNum); // Mark the instruction as squashed, and ready to commit so that diff --git a/src/cpu/o3/thread_context.hh b/src/cpu/o3/thread_context.hh index b7790cfda..4e559000b 100755 --- a/src/cpu/o3/thread_context.hh +++ b/src/cpu/o3/thread_context.hh @@ -172,29 +172,24 @@ class O3ThreadContext : public ThreadContext virtual void setFloatRegBits(int reg_idx, FloatRegBits val); - /** Reads this thread's PC. */ - virtual uint64_t readPC() - { return cpu->readPC(thread->threadId()); } - - /** Sets this thread's PC. */ - virtual void setPC(uint64_t val); - - /** Reads this thread's next PC. */ - virtual uint64_t readNextPC() - { return cpu->readNextPC(thread->threadId()); } + /** Reads this thread's PC state. */ + virtual TheISA::PCState pcState() + { return cpu->pcState(thread->threadId()); } - /** Sets this thread's next PC. */ - virtual void setNextPC(uint64_t val); + /** Sets this thread's PC state. */ + virtual void pcState(const TheISA::PCState &val); - virtual uint64_t readMicroPC() - { return cpu->readMicroPC(thread->threadId()); } - - virtual void setMicroPC(uint64_t val); + /** Reads this thread's PC. */ + virtual Addr instAddr() + { return cpu->instAddr(thread->threadId()); } - virtual uint64_t readNextMicroPC() - { return cpu->readNextMicroPC(thread->threadId()); } + /** Reads this thread's next PC. */ + virtual Addr nextInstAddr() + { return cpu->nextInstAddr(thread->threadId()); } - virtual void setNextMicroPC(uint64_t val); + /** Reads this thread's next PC. */ + virtual MicroPC microPC() + { return cpu->microPC(thread->threadId()); } /** Reads a miscellaneous register. */ virtual MiscReg readMiscRegNoEffect(int misc_reg) @@ -247,15 +242,6 @@ class O3ThreadContext : public ThreadContext } #endif - virtual uint64_t readNextNPC() - { - return this->cpu->readNextNPC(this->thread->threadId()); - } - - virtual void setNextNPC(uint64_t val) - { - this->cpu->setNextNPC(val, this->thread->threadId()); - } }; #endif diff --git a/src/cpu/o3/thread_context_impl.hh b/src/cpu/o3/thread_context_impl.hh index 0b7a2b172..367442830 100755 --- a/src/cpu/o3/thread_context_impl.hh +++ b/src/cpu/o3/thread_context_impl.hh @@ -248,11 +248,7 @@ O3ThreadContext<Impl>::copyArchRegs(ThreadContext *tc) // Then finally set the PC, the next PC, the nextNPC, the micropc, and the // next micropc. - cpu->setPC(tc->readPC(), tid); - cpu->setNextPC(tc->readNextPC(), tid); - cpu->setNextNPC(tc->readNextNPC(), tid); - cpu->setMicroPC(tc->readMicroPC(), tid); - cpu->setNextMicroPC(tc->readNextMicroPC(), tid); + cpu->pcState(tc->pcState(), tid); #if !FULL_SYSTEM this->thread->funcExeInst = tc->readFuncExeInst(); #endif @@ -327,45 +323,9 @@ O3ThreadContext<Impl>::setFloatRegBits(int reg_idx, FloatRegBits val) template <class Impl> void -O3ThreadContext<Impl>::setPC(uint64_t val) +O3ThreadContext<Impl>::pcState(const TheISA::PCState &val) { - cpu->setPC(val, thread->threadId()); - - // Squash if we're not already in a state update mode. - if (!thread->trapPending && !thread->inSyscall) { - cpu->squashFromTC(thread->threadId()); - } -} - -template <class Impl> -void -O3ThreadContext<Impl>::setNextPC(uint64_t val) -{ - cpu->setNextPC(val, thread->threadId()); - - // Squash if we're not already in a state update mode. - if (!thread->trapPending && !thread->inSyscall) { - cpu->squashFromTC(thread->threadId()); - } -} - -template <class Impl> -void -O3ThreadContext<Impl>::setMicroPC(uint64_t val) -{ - cpu->setMicroPC(val, thread->threadId()); - - // Squash if we're not already in a state update mode. - if (!thread->trapPending && !thread->inSyscall) { - cpu->squashFromTC(thread->threadId()); - } -} - -template <class Impl> -void -O3ThreadContext<Impl>::setNextMicroPC(uint64_t val) -{ - cpu->setNextMicroPC(val, thread->threadId()); + cpu->pcState(val, thread->threadId()); // Squash if we're not already in a state update mode. if (!thread->trapPending && !thread->inSyscall) { diff --git a/src/cpu/pc_event.cc b/src/cpu/pc_event.cc index 79f5277d5..533d61498 100644 --- a/src/cpu/pc_event.cc +++ b/src/cpu/pc_event.cc @@ -83,7 +83,7 @@ PCEventQueue::schedule(PCEvent *event) bool PCEventQueue::doService(ThreadContext *tc) { - Addr pc = tc->readPC() & ~0x3; + Addr pc = tc->instAddr() & ~0x3; int serviced = 0; range_t range = equal_range(pc); for (iterator i = range.first; i != range.second; ++i) { @@ -91,7 +91,7 @@ PCEventQueue::doService(ThreadContext *tc) // another event. This for example, prevents two invocations // of the SkipFuncEvent. Maybe we should have separate PC // event queues for each processor? - if (pc != (tc->readPC() & ~0x3)) + if (pc != (tc->instAddr() & ~0x3)) continue; DPRINTF(PCEvent, "PC based event serviced at %#x: %s\n", diff --git a/src/cpu/pred/btb.cc b/src/cpu/pred/btb.cc index c6a5e23f9..e87cc6dc2 100644 --- a/src/cpu/pred/btb.cc +++ b/src/cpu/pred/btb.cc @@ -68,25 +68,25 @@ DefaultBTB::reset() inline unsigned -DefaultBTB::getIndex(const Addr &inst_PC) +DefaultBTB::getIndex(Addr instPC) { // Need to shift PC over by the word offset. - return (inst_PC >> instShiftAmt) & idxMask; + return (instPC >> instShiftAmt) & idxMask; } inline Addr -DefaultBTB::getTag(const Addr &inst_PC) +DefaultBTB::getTag(Addr instPC) { - return (inst_PC >> tagShiftAmt) & tagMask; + return (instPC >> tagShiftAmt) & tagMask; } bool -DefaultBTB::valid(const Addr &inst_PC, ThreadID tid) +DefaultBTB::valid(Addr instPC, ThreadID tid) { - unsigned btb_idx = getIndex(inst_PC); + unsigned btb_idx = getIndex(instPC); - Addr inst_tag = getTag(inst_PC); + Addr inst_tag = getTag(instPC); assert(btb_idx < numEntries); @@ -102,12 +102,12 @@ DefaultBTB::valid(const Addr &inst_PC, ThreadID tid) // @todo Create some sort of return struct that has both whether or not the // address is valid, and also the address. For now will just use addr = 0 to // represent invalid entry. -Addr -DefaultBTB::lookup(const Addr &inst_PC, ThreadID tid) +TheISA::PCState +DefaultBTB::lookup(Addr instPC, ThreadID tid) { - unsigned btb_idx = getIndex(inst_PC); + unsigned btb_idx = getIndex(instPC); - Addr inst_tag = getTag(inst_PC); + Addr inst_tag = getTag(instPC); assert(btb_idx < numEntries); @@ -121,14 +121,14 @@ DefaultBTB::lookup(const Addr &inst_PC, ThreadID tid) } void -DefaultBTB::update(const Addr &inst_PC, const Addr &target, ThreadID tid) +DefaultBTB::update(Addr instPC, const TheISA::PCState &target, ThreadID tid) { - unsigned btb_idx = getIndex(inst_PC); + unsigned btb_idx = getIndex(instPC); assert(btb_idx < numEntries); btb[btb_idx].tid = tid; btb[btb_idx].valid = true; btb[btb_idx].target = target; - btb[btb_idx].tag = getTag(inst_PC); + btb[btb_idx].tag = getTag(instPC); } diff --git a/src/cpu/pred/btb.hh b/src/cpu/pred/btb.hh index 6557522e0..814b23872 100644 --- a/src/cpu/pred/btb.hh +++ b/src/cpu/pred/btb.hh @@ -31,8 +31,10 @@ #ifndef __CPU_O3_BTB_HH__ #define __CPU_O3_BTB_HH__ +#include "arch/types.hh" #include "base/misc.hh" #include "base/types.hh" +#include "config/the_isa.hh" class DefaultBTB { @@ -41,14 +43,13 @@ class DefaultBTB { BTBEntry() : tag(0), target(0), valid(false) - { - } + {} /** The entry's tag. */ Addr tag; /** The entry's target. */ - Addr target; + TheISA::PCState target; /** The entry's thread id. */ ThreadID tid; @@ -74,21 +75,21 @@ class DefaultBTB * @param tid The thread id. * @return Returns the target of the branch. */ - Addr lookup(const Addr &inst_PC, ThreadID tid); + TheISA::PCState lookup(Addr instPC, ThreadID tid); /** Checks if a branch is in the BTB. * @param inst_PC The address of the branch to look up. * @param tid The thread id. * @return Whether or not the branch exists in the BTB. */ - bool valid(const Addr &inst_PC, ThreadID tid); + bool valid(Addr instPC, ThreadID tid); /** Updates the BTB with the target of a branch. * @param inst_PC The address of the branch being updated. * @param target_PC The target address of the branch. * @param tid The thread id. */ - void update(const Addr &inst_PC, const Addr &target_PC, + void update(Addr instPC, const TheISA::PCState &targetPC, ThreadID tid); private: @@ -96,13 +97,13 @@ class DefaultBTB * @param inst_PC The branch to look up. * @return Returns the index into the BTB. */ - inline unsigned getIndex(const Addr &inst_PC); + inline unsigned getIndex(Addr instPC); /** Returns the tag bits of a given address. * @param inst_PC The branch's address. * @return Returns the tag bits. */ - inline Addr getTag(const Addr &inst_PC); + inline Addr getTag(Addr instPC); /** The actual BTB. */ std::vector<BTBEntry> btb; diff --git a/src/cpu/pred/ras.cc b/src/cpu/pred/ras.cc index 6373e5de3..0ba09bfae 100644 --- a/src/cpu/pred/ras.cc +++ b/src/cpu/pred/ras.cc @@ -34,13 +34,8 @@ void ReturnAddrStack::init(unsigned _numEntries) { numEntries = _numEntries; - usedEntries = 0; - tos = 0; - addrStack.resize(numEntries); - - for (unsigned i = 0; i < numEntries; ++i) - addrStack[i] = 0; + reset(); } void @@ -49,11 +44,11 @@ ReturnAddrStack::reset() usedEntries = 0; tos = 0; for (unsigned i = 0; i < numEntries; ++i) - addrStack[i] = 0; + addrStack[i].set(0); } void -ReturnAddrStack::push(const Addr &return_addr) +ReturnAddrStack::push(const TheISA::PCState &return_addr) { incrTos(); @@ -76,9 +71,9 @@ ReturnAddrStack::pop() void ReturnAddrStack::restore(unsigned top_entry_idx, - const Addr &restored_target) + const TheISA::PCState &restored) { tos = top_entry_idx; - addrStack[tos] = restored_target; + addrStack[tos] = restored; } diff --git a/src/cpu/pred/ras.hh b/src/cpu/pred/ras.hh index a36faf79a..ab92b34c2 100644 --- a/src/cpu/pred/ras.hh +++ b/src/cpu/pred/ras.hh @@ -33,7 +33,9 @@ #include <vector> +#include "arch/types.hh" #include "base/types.hh" +#include "config/the_isa.hh" /** Return address stack class, implements a simple RAS. */ class ReturnAddrStack @@ -52,7 +54,7 @@ class ReturnAddrStack void reset(); /** Returns the top address on the RAS. */ - Addr top() + TheISA::PCState top() { return addrStack[tos]; } /** Returns the index of the top of the RAS. */ @@ -60,7 +62,7 @@ class ReturnAddrStack { return tos; } /** Pushes an address onto the RAS. */ - void push(const Addr &return_addr); + void push(const TheISA::PCState &return_addr); /** Pops the top address from the RAS. */ void pop(); @@ -68,9 +70,9 @@ class ReturnAddrStack /** Changes index to the top of the RAS, and replaces the top address with * a new target. * @param top_entry_idx The index of the RAS that will now be the top. - * @param restored_target The new target address of the new top of the RAS. + * @param restored The new target address of the new top of the RAS. */ - void restore(unsigned top_entry_idx, const Addr &restored_target); + void restore(unsigned top_entry_idx, const TheISA::PCState &restored); bool empty() { return usedEntries == 0; } @@ -85,7 +87,7 @@ class ReturnAddrStack { tos = (tos == 0 ? numEntries - 1 : tos - 1); } /** The RAS itself. */ - std::vector<Addr> addrStack; + std::vector<TheISA::PCState> addrStack; /** The number of entries in the RAS. */ unsigned numEntries; diff --git a/src/cpu/simple/atomic.cc b/src/cpu/simple/atomic.cc index d97e7aeec..de26ca2f8 100644 --- a/src/cpu/simple/atomic.cc +++ b/src/cpu/simple/atomic.cc @@ -321,7 +321,7 @@ AtomicSimpleCPU::readBytes(Addr addr, uint8_t * data, dcache_latency = 0; while (1) { - req->setVirt(0, addr, size, flags, thread->readPC()); + req->setVirt(0, addr, size, flags, thread->pcState().instAddr()); // translate to physical address Fault fault = thread->dtb->translateAtomic(req, tc, BaseTLB::Read); @@ -475,7 +475,7 @@ AtomicSimpleCPU::writeBytes(uint8_t *data, unsigned size, dcache_latency = 0; while(1) { - req->setVirt(0, addr, size, flags, thread->readPC()); + req->setVirt(0, addr, size, flags, thread->pcState().instAddr()); // translate to physical address Fault fault = thread->dtb->translateAtomic(req, tc, BaseTLB::Write); @@ -643,8 +643,11 @@ AtomicSimpleCPU::tick() Fault fault = NoFault; - bool fromRom = isRomMicroPC(thread->readMicroPC()); - if (!fromRom && !curMacroStaticInst) { + TheISA::PCState pcState = thread->pcState(); + + bool needToFetch = !isRomMicroPC(pcState.microPC()) && + !curMacroStaticInst; + if (needToFetch) { setupFetchRequest(&ifetch_req); fault = thread->itb->translateAtomic(&ifetch_req, tc, BaseTLB::Execute); @@ -655,7 +658,7 @@ AtomicSimpleCPU::tick() bool icache_access = false; dcache_access = false; // assume no dcache access - if (!fromRom && !curMacroStaticInst) { + if (needToFetch) { // This is commented out because the predecoder would act like // a tiny cache otherwise. It wouldn't be flushed when needed // like the I cache. It should be flushed, and when that works diff --git a/src/cpu/simple/base.cc b/src/cpu/simple/base.cc index 088d5ff16..196b72cc0 100644 --- a/src/cpu/simple/base.cc +++ b/src/cpu/simple/base.cc @@ -368,19 +368,13 @@ BaseSimpleCPU::checkForInterrupts() void BaseSimpleCPU::setupFetchRequest(Request *req) { - Addr threadPC = thread->readPC(); + Addr instAddr = thread->instAddr(); // set up memory request for instruction fetch -#if ISA_HAS_DELAY_SLOT - DPRINTF(Fetch,"Fetch: PC:%08p NPC:%08p NNPC:%08p\n",threadPC, - thread->readNextPC(),thread->readNextNPC()); -#else - DPRINTF(Fetch,"Fetch: PC:%08p NPC:%08p\n",threadPC, - thread->readNextPC()); -#endif + DPRINTF(Fetch, "Fetch: PC:%08p\n", instAddr); - Addr fetchPC = (threadPC & PCMask) + fetchOffset; - req->setVirt(0, fetchPC, sizeof(MachInst), Request::INST_FETCH, threadPC); + Addr fetchPC = (instAddr & PCMask) + fetchOffset; + req->setVirt(0, fetchPC, sizeof(MachInst), Request::INST_FETCH, instAddr); } @@ -399,11 +393,12 @@ BaseSimpleCPU::preExecute() // decode the instruction inst = gtoh(inst); - MicroPC upc = thread->readMicroPC(); + TheISA::PCState pcState = thread->pcState(); - if (isRomMicroPC(upc)) { + if (isRomMicroPC(pcState.microPC())) { stayAtPC = false; - curStaticInst = microcodeRom.fetchMicroop(upc, curMacroStaticInst); + curStaticInst = microcodeRom.fetchMicroop(pcState.microPC(), + curMacroStaticInst); } else if (!curMacroStaticInst) { //We're not in the middle of a macro instruction StaticInstPtr instPtr = NULL; @@ -412,21 +407,19 @@ BaseSimpleCPU::preExecute() //This should go away once the constructor can be set up properly predecoder.setTC(thread->getTC()); //If more fetch data is needed, pass it in. - Addr fetchPC = (thread->readPC() & PCMask) + fetchOffset; + Addr fetchPC = (pcState.instAddr() & PCMask) + fetchOffset; //if(predecoder.needMoreBytes()) - predecoder.moreBytes(thread->readPC(), fetchPC, inst); + predecoder.moreBytes(pcState, fetchPC, inst); //else // predecoder.process(); //If an instruction is ready, decode it. Otherwise, we'll have to //fetch beyond the MachInst at the current pc. if (predecoder.extMachInstReady()) { -#if THE_ISA == X86_ISA || THE_ISA == ARM_ISA - thread->setNextPC(thread->readPC() + predecoder.getInstSize()); -#endif // X86_ISA stayAtPC = false; - instPtr = StaticInst::decode(predecoder.getExtMachInst(), - thread->readPC()); + ExtMachInst machInst = predecoder.getExtMachInst(pcState); + thread->pcState(pcState); + instPtr = StaticInst::decode(machInst, pcState.instAddr()); } else { stayAtPC = true; fetchOffset += sizeof(MachInst); @@ -436,13 +429,13 @@ BaseSimpleCPU::preExecute() //out micro ops if (instPtr && instPtr->isMacroop()) { curMacroStaticInst = instPtr; - curStaticInst = curMacroStaticInst->fetchMicroop(upc); + curStaticInst = curMacroStaticInst->fetchMicroop(pcState.microPC()); } else { curStaticInst = instPtr; } } else { //Read the next micro op from the macro op - curStaticInst = curMacroStaticInst->fetchMicroop(upc); + curStaticInst = curMacroStaticInst->fetchMicroop(pcState.microPC()); } //If we decoded an instruction this "tick", record information about it. @@ -450,8 +443,7 @@ BaseSimpleCPU::preExecute() { #if TRACING_ON traceData = tracer->getInstRecord(curTick, tc, - curStaticInst, thread->readPC(), - curMacroStaticInst, thread->readMicroPC()); + curStaticInst, thread->pcState(), curMacroStaticInst); DPRINTF(Decode,"Decode: Decoded %s instruction: 0x%x\n", curStaticInst->getName(), curStaticInst->machInst); @@ -462,10 +454,14 @@ BaseSimpleCPU::preExecute() void BaseSimpleCPU::postExecute() { + assert(curStaticInst); + + TheISA::PCState pc = tc->pcState(); + Addr instAddr = pc.instAddr(); #if FULL_SYSTEM - if (thread->profile && curStaticInst) { + if (thread->profile) { bool usermode = TheISA::inUserMode(tc); - thread->profilePC = usermode ? 1 : thread->readPC(); + thread->profilePC = usermode ? 1 : instAddr; ProfileNode *node = thread->profile->consume(tc, curStaticInst); if (node) thread->profileNode = node; @@ -482,10 +478,10 @@ BaseSimpleCPU::postExecute() } if (CPA::available()) { - CPA::cpa()->swAutoBegin(tc, thread->readNextPC()); + CPA::cpa()->swAutoBegin(tc, pc.nextInstAddr()); } - traceFunctions(thread->readPC()); + traceFunctions(instAddr); if (traceData) { traceData->dump(); @@ -505,30 +501,12 @@ BaseSimpleCPU::advancePC(Fault fault) fault->invoke(tc, curStaticInst); predecoder.reset(); } else { - //If we're at the last micro op for this instruction - if (curStaticInst && curStaticInst->isLastMicroop()) { - //We should be working with a macro op or be in the ROM - assert(curMacroStaticInst || - isRomMicroPC(thread->readMicroPC())); - //Close out this macro op, and clean up the - //microcode state - curMacroStaticInst = StaticInst::nullStaticInstPtr; - thread->setMicroPC(normalMicroPC(0)); - thread->setNextMicroPC(normalMicroPC(1)); - } - //If we're still in a macro op - if (curMacroStaticInst || isRomMicroPC(thread->readMicroPC())) { - //Advance the micro pc - thread->setMicroPC(thread->readNextMicroPC()); - //Advance the "next" micro pc. Note that there are no delay - //slots, and micro ops are "word" addressed. - thread->setNextMicroPC(thread->readNextMicroPC() + 1); - } else { - // go to the next instruction - thread->setPC(thread->readNextPC()); - thread->setNextPC(thread->readNextNPC()); - thread->setNextNPC(thread->readNextNPC() + sizeof(MachInst)); - assert(thread->readNextPC() != thread->readNextNPC()); + if (curStaticInst) { + if (curStaticInst->isLastMicroop()) + curMacroStaticInst = StaticInst::nullStaticInstPtr; + TheISA::PCState pcState = thread->pcState(); + TheISA::advancePC(pcState, curStaticInst); + thread->pcState(pcState); } } } diff --git a/src/cpu/simple/base.hh b/src/cpu/simple/base.hh index 0aa0a295c..a713533fc 100644 --- a/src/cpu/simple/base.hh +++ b/src/cpu/simple/base.hh @@ -88,11 +88,12 @@ class BaseSimpleCPU : public BaseCPU Trace::InstRecord *traceData; inline void checkPcEventQueue() { - Addr oldpc; + Addr oldpc, pc = thread->instAddr(); do { - oldpc = thread->readPC(); + oldpc = pc; system->pcEventQueue.service(tc); - } while (oldpc != thread->readPC()); + pc = thread->instAddr(); + } while (oldpc != pc); } public: @@ -282,18 +283,7 @@ class BaseSimpleCPU : public BaseCPU thread->setFloatRegBits(reg_idx, val); } - uint64_t readPC() { return thread->readPC(); } - uint64_t readMicroPC() { return thread->readMicroPC(); } - uint64_t readNextPC() { return thread->readNextPC(); } - uint64_t readNextMicroPC() { return thread->readNextMicroPC(); } - uint64_t readNextNPC() { return thread->readNextNPC(); } bool readPredicate() { return thread->readPredicate(); } - - void setPC(uint64_t val) { thread->setPC(val); } - void setMicroPC(uint64_t val) { thread->setMicroPC(val); } - void setNextPC(uint64_t val) { thread->setNextPC(val); } - void setNextMicroPC(uint64_t val) { thread->setNextMicroPC(val); } - void setNextNPC(uint64_t val) { thread->setNextNPC(val); } void setPredicate(bool val) { thread->setPredicate(val); @@ -301,6 +291,11 @@ class BaseSimpleCPU : public BaseCPU traceData->setPredicate(val); } } + TheISA::PCState pcState() { return thread->pcState(); } + void pcState(const TheISA::PCState &val) { thread->pcState(val); } + Addr instAddr() { return thread->instAddr(); } + Addr nextInstAddr() { return thread->nextInstAddr(); } + MicroPC microPC() { return thread->microPC(); } MiscReg readMiscRegNoEffect(int misc_reg) { diff --git a/src/cpu/simple/timing.cc b/src/cpu/simple/timing.cc index 2eb5b432b..863c28be2 100644 --- a/src/cpu/simple/timing.cc +++ b/src/cpu/simple/timing.cc @@ -422,7 +422,7 @@ TimingSimpleCPU::readBytes(Addr addr, uint8_t *data, Fault fault; const int asid = 0; const ThreadID tid = 0; - const Addr pc = thread->readPC(); + const Addr pc = thread->instAddr(); unsigned block_size = dcachePort.peerBlockSize(); BaseTLB::Mode mode = BaseTLB::Read; @@ -545,7 +545,7 @@ TimingSimpleCPU::writeTheseBytes(uint8_t *data, unsigned size, { const int asid = 0; const ThreadID tid = 0; - const Addr pc = thread->readPC(); + const Addr pc = thread->instAddr(); unsigned block_size = dcachePort.peerBlockSize(); BaseTLB::Mode mode = BaseTLB::Write; @@ -701,9 +701,10 @@ TimingSimpleCPU::fetch() checkPcEventQueue(); - bool fromRom = isRomMicroPC(thread->readMicroPC()); + TheISA::PCState pcState = thread->pcState(); + bool needToFetch = !isRomMicroPC(pcState.microPC()) && !curMacroStaticInst; - if (!fromRom && !curMacroStaticInst) { + if (needToFetch) { Request *ifetch_req = new Request(); ifetch_req->setThreadContext(_cpuId, /* thread ID */ 0); setupFetchRequest(ifetch_req); diff --git a/src/cpu/simple_thread.cc b/src/cpu/simple_thread.cc index bd796428a..15fac8677 100644 --- a/src/cpu/simple_thread.cc +++ b/src/cpu/simple_thread.cc @@ -193,11 +193,7 @@ SimpleThread::serialize(ostream &os) ThreadState::serialize(os); SERIALIZE_ARRAY(floatRegs.i, TheISA::NumFloatRegs); SERIALIZE_ARRAY(intRegs, TheISA::NumIntRegs); - SERIALIZE_SCALAR(microPC); - SERIALIZE_SCALAR(nextMicroPC); - SERIALIZE_SCALAR(PC); - SERIALIZE_SCALAR(nextPC); - SERIALIZE_SCALAR(nextNPC); + _pcState.serialize(os); // thread_num and cpu_id are deterministic from the config // @@ -213,11 +209,7 @@ SimpleThread::unserialize(Checkpoint *cp, const std::string §ion) ThreadState::unserialize(cp, section); UNSERIALIZE_ARRAY(floatRegs.i, TheISA::NumFloatRegs); UNSERIALIZE_ARRAY(intRegs, TheISA::NumIntRegs); - UNSERIALIZE_SCALAR(microPC); - UNSERIALIZE_SCALAR(nextMicroPC); - UNSERIALIZE_SCALAR(PC); - UNSERIALIZE_SCALAR(nextPC); - UNSERIALIZE_SCALAR(nextNPC); + _pcState.unserialize(cp, section); // thread_num and cpu_id are deterministic from the config // diff --git a/src/cpu/simple_thread.hh b/src/cpu/simple_thread.hh index e4a7b7a77..48077a9b9 100644 --- a/src/cpu/simple_thread.hh +++ b/src/cpu/simple_thread.hh @@ -106,27 +106,7 @@ class SimpleThread : public ThreadState TheISA::IntReg intRegs[TheISA::NumIntRegs]; TheISA::ISA isa; // one "instance" of the current ISA. - /** The current microcode pc for the currently executing macro - * operation. - */ - MicroPC microPC; - - /** The next microcode pc for the currently executing macro - * operation. - */ - MicroPC nextMicroPC; - - /** The current pc. - */ - Addr PC; - - /** The next pc. - */ - Addr nextPC; - - /** The next next pc. - */ - Addr nextNPC; + TheISA::PCState _pcState; /** Did this instruction execute or is it predicated false */ bool predicate; @@ -245,9 +225,7 @@ class SimpleThread : public ThreadState void clearArchRegs() { - microPC = 0; - nextMicroPC = 1; - PC = nextPC = nextNPC = 0; + _pcState = 0; memset(intRegs, 0, sizeof(intRegs)); memset(floatRegs.i, 0, sizeof(floatRegs.i)); isa.clear(); @@ -313,60 +291,34 @@ class SimpleThread : public ThreadState reg_idx, flatIndex, val, floatRegs.f[flatIndex]); } - uint64_t readPC() - { - return PC; - } - - void setPC(uint64_t val) - { - PC = val; - } - - uint64_t readMicroPC() - { - return microPC; - } - - void setMicroPC(uint64_t val) - { - microPC = val; - } - - uint64_t readNextPC() + TheISA::PCState + pcState() { - return nextPC; + return _pcState; } - void setNextPC(uint64_t val) + void + pcState(const TheISA::PCState &val) { - nextPC = val; + _pcState = val; } - uint64_t readNextMicroPC() + Addr + instAddr() { - return nextMicroPC; + return _pcState.instAddr(); } - void setNextMicroPC(uint64_t val) + Addr + nextInstAddr() { - nextMicroPC = val; + return _pcState.nextInstAddr(); } - uint64_t readNextNPC() + MicroPC + microPC() { -#if ISA_HAS_DELAY_SLOT - return nextNPC; -#else - return nextPC + sizeof(TheISA::MachInst); -#endif - } - - void setNextNPC(uint64_t val) - { -#if ISA_HAS_DELAY_SLOT - nextNPC = val; -#endif + return _pcState.microPC(); } bool readPredicate() diff --git a/src/cpu/static_inst.cc b/src/cpu/static_inst.cc index 2c4fc8ab9..f2a72c96a 100644 --- a/src/cpu/static_inst.cc +++ b/src/cpu/static_inst.cc @@ -68,7 +68,8 @@ StaticInst::dumpDecodeCacheStats() } bool -StaticInst::hasBranchTarget(Addr pc, ThreadContext *tc, Addr &tgt) const +StaticInst::hasBranchTarget(const TheISA::PCState &pc, ThreadContext *tc, + TheISA::PCState &tgt) const { if (isDirectCtrl()) { tgt = branchTarget(pc); @@ -84,21 +85,21 @@ StaticInst::hasBranchTarget(Addr pc, ThreadContext *tc, Addr &tgt) const } StaticInstPtr -StaticInst::fetchMicroop(MicroPC micropc) +StaticInst::fetchMicroop(MicroPC upc) const { panic("StaticInst::fetchMicroop() called on instruction " "that is not microcoded."); } -Addr -StaticInst::branchTarget(Addr branchPC) const +TheISA::PCState +StaticInst::branchTarget(const TheISA::PCState &pc) const { panic("StaticInst::branchTarget() called on instruction " "that is not a PC-relative branch."); M5_DUMMY_RETURN; } -Addr +TheISA::PCState StaticInst::branchTarget(ThreadContext *tc) const { panic("StaticInst::branchTarget() called on instruction " diff --git a/src/cpu/static_inst.hh b/src/cpu/static_inst.hh index 6474bbf9c..b219fafd6 100644 --- a/src/cpu/static_inst.hh +++ b/src/cpu/static_inst.hh @@ -35,6 +35,7 @@ #include <string> #include "arch/isa_traits.hh" +#include "arch/types.hh" #include "arch/registers.hh" #include "config/the_isa.hh" #include "base/hashmap.hh" @@ -70,28 +71,6 @@ namespace Trace { class InstRecord; } -typedef uint16_t MicroPC; - -static const MicroPC MicroPCRomBit = 1 << (sizeof(MicroPC) * 8 - 1); - -static inline MicroPC -romMicroPC(MicroPC upc) -{ - return upc | MicroPCRomBit; -} - -static inline MicroPC -normalMicroPC(MicroPC upc) -{ - return upc & ~MicroPCRomBit; -} - -static inline bool -isRomMicroPC(MicroPC upc) -{ - return MicroPCRomBit & upc; -} - /** * Base, ISA-independent static instruction class. * @@ -392,18 +371,20 @@ class StaticInst : public StaticInstBase */ #include "cpu/static_inst_exec_sigs.hh" + virtual void advancePC(TheISA::PCState &pcState) const = 0; + /** * Return the microop that goes with a particular micropc. This should * only be defined/used in macroops which will contain microops */ - virtual StaticInstPtr fetchMicroop(MicroPC micropc); + virtual StaticInstPtr fetchMicroop(MicroPC upc) const; /** * Return the target address for a PC-relative branch. * Invalid if not a PC-relative branch (i.e. isDirectCtrl() * should be true). */ - virtual Addr branchTarget(Addr branchPC) const; + virtual TheISA::PCState branchTarget(const TheISA::PCState &pc) const; /** * Return the target address for an indirect branch (jump). The @@ -412,13 +393,14 @@ class StaticInst : public StaticInstBase * execute the branch in question. Invalid if not an indirect * branch (i.e. isIndirectCtrl() should be true). */ - virtual Addr branchTarget(ThreadContext *tc) const; + virtual TheISA::PCState branchTarget(ThreadContext *tc) const; /** * Return true if the instruction is a control transfer, and if so, * return the target address as well. */ - bool hasBranchTarget(Addr pc, ThreadContext *tc, Addr &tgt) const; + bool hasBranchTarget(const TheISA::PCState &pc, ThreadContext *tc, + TheISA::PCState &tgt) const; /** * Return string representation of disassembled instruction. diff --git a/src/cpu/thread_context.cc b/src/cpu/thread_context.cc index f2083ef08..c4960ea30 100644 --- a/src/cpu/thread_context.cc +++ b/src/cpu/thread_context.cc @@ -65,16 +65,8 @@ ThreadContext::compare(ThreadContext *one, ThreadContext *two) } #endif - Addr pc1 = one->readPC(); - Addr pc2 = two->readPC(); - if (pc1 != pc2) - panic("PCs doesn't match, one: %#x, two: %#x", pc1, pc2); - - Addr npc1 = one->readNextPC(); - Addr npc2 = two->readNextPC(); - if (npc1 != npc2) - panic("NPCs doesn't match, one: %#x, two: %#x", npc1, npc2); - + if (!(one->pcState() == two->pcState())) + panic("PC state doesn't match."); int id1 = one->cpuId(); int id2 = two->cpuId(); if (id1 != id2) diff --git a/src/cpu/thread_context.hh b/src/cpu/thread_context.hh index 753fa2146..1c70ef59a 100644 --- a/src/cpu/thread_context.hh +++ b/src/cpu/thread_context.hh @@ -192,25 +192,15 @@ class ThreadContext virtual void setFloatRegBits(int reg_idx, FloatRegBits val) = 0; - virtual uint64_t readPC() = 0; + virtual TheISA::PCState pcState() = 0; - virtual void setPC(uint64_t val) = 0; + virtual void pcState(const TheISA::PCState &val) = 0; - virtual uint64_t readNextPC() = 0; + virtual Addr instAddr() = 0; - virtual void setNextPC(uint64_t val) = 0; + virtual Addr nextInstAddr() = 0; - virtual uint64_t readNextNPC() = 0; - - virtual void setNextNPC(uint64_t val) = 0; - - virtual uint64_t readMicroPC() = 0; - - virtual void setMicroPC(uint64_t val) = 0; - - virtual uint64_t readNextMicroPC() = 0; - - virtual void setNextMicroPC(uint64_t val) = 0; + virtual MicroPC microPC() = 0; virtual MiscReg readMiscRegNoEffect(int misc_reg) = 0; @@ -377,25 +367,13 @@ class ProxyThreadContext : public ThreadContext void setFloatRegBits(int reg_idx, FloatRegBits val) { actualTC->setFloatRegBits(reg_idx, val); } - uint64_t readPC() { return actualTC->readPC(); } - - void setPC(uint64_t val) { actualTC->setPC(val); } - - uint64_t readNextPC() { return actualTC->readNextPC(); } - - void setNextPC(uint64_t val) { actualTC->setNextPC(val); } - - uint64_t readNextNPC() { return actualTC->readNextNPC(); } - - void setNextNPC(uint64_t val) { actualTC->setNextNPC(val); } - - uint64_t readMicroPC() { return actualTC->readMicroPC(); } - - void setMicroPC(uint64_t val) { actualTC->setMicroPC(val); } + TheISA::PCState pcState() { return actualTC->pcState(); } - uint64_t readNextMicroPC() { return actualTC->readMicroPC(); } + void pcState(const TheISA::PCState &val) { actualTC->pcState(val); } - void setNextMicroPC(uint64_t val) { actualTC->setNextMicroPC(val); } + Addr instAddr() { return actualTC->instAddr(); } + Addr nextInstAddr() { return actualTC->nextInstAddr(); } + MicroPC microPC() { return actualTC->microPC(); } bool readPredicate() { return actualTC->readPredicate(); } |