From 6f4bd2c1da0dc7783da87c4877a41332901199b2 Mon Sep 17 00:00:00 2001 From: Gabe Black Date: Sun, 31 Oct 2010 00:07:20 -0700 Subject: [PATCH] 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. --- src/arch/alpha/ev5.cc | 11 +- src/arch/alpha/faults.cc | 10 +- src/arch/alpha/interrupts.hh | 2 +- src/arch/alpha/isa/branch.isa | 40 +- src/arch/alpha/isa/decoder.isa | 14 +- src/arch/alpha/isa/main.isa | 9 +- src/arch/alpha/predecoder.hh | 6 +- src/arch/alpha/process.cc | 10 +- src/arch/alpha/remote_gdb.cc | 20 +- src/arch/alpha/stacktrace.cc | 4 +- src/arch/alpha/tlb.cc | 4 +- src/arch/alpha/types.hh | 3 + src/arch/alpha/utility.cc | 9 +- src/arch/alpha/utility.hh | 16 + src/arch/arm/faults.cc | 34 +- src/arch/arm/insts/macromem.hh | 12 + src/arch/arm/insts/mem.hh | 38 +- src/arch/arm/insts/pred_inst.hh | 11 +- src/arch/arm/insts/static_inst.hh | 62 +-- src/arch/arm/insts/vfp.hh | 12 + src/arch/arm/isa.cc | 25 +- src/arch/arm/isa/formats/breakpoint.isa | 2 +- src/arch/arm/isa/insts/branch.isa | 56 ++- src/arch/arm/isa/insts/data.isa | 7 +- src/arch/arm/isa/insts/ldr.isa | 6 +- src/arch/arm/isa/insts/macromem.isa | 4 +- src/arch/arm/isa/insts/misc.isa | 24 +- src/arch/arm/isa/operands.isa | 204 ++++----- src/arch/arm/isa_traits.hh | 9 - src/arch/arm/linux/system.cc | 3 +- src/arch/arm/nativetrace.cc | 4 +- src/arch/arm/predecoder.cc | 6 +- src/arch/arm/predecoder.hh | 5 +- src/arch/arm/process.cc | 10 +- src/arch/arm/system.hh | 2 +- src/arch/arm/table_walker.cc | 2 +- src/arch/arm/tlb.cc | 4 +- src/arch/arm/types.hh | 185 ++++++++ src/arch/arm/utility.cc | 10 +- src/arch/arm/utility.hh | 21 +- src/arch/generic/types.hh | 432 ++++++++++++++++++ src/arch/isa_parser.py | 54 +-- src/arch/mips/isa/base.isa | 7 + src/arch/mips/isa/decoder.isa | 44 +- src/arch/mips/isa/formats/branch.isa | 90 ++-- src/arch/mips/isa/includes.isa | 1 + src/arch/mips/isa/operands.isa | 3 +- src/arch/mips/mt.hh | 14 +- src/arch/mips/predecoder.hh | 4 +- src/arch/mips/process.cc | 5 +- src/arch/mips/types.hh | 3 + src/arch/mips/utility.cc | 7 +- src/arch/mips/utility.hh | 16 + src/arch/power/insts/branch.cc | 24 +- src/arch/power/insts/branch.hh | 10 +- src/arch/power/insts/static_inst.hh | 6 + src/arch/power/isa/decoder.isa | 36 +- src/arch/power/isa/formats/branch.isa | 6 +- src/arch/power/isa/formats/unknown.isa | 2 +- src/arch/power/isa/operands.isa | 3 +- src/arch/power/predecoder.hh | 4 +- src/arch/power/process.cc | 4 +- src/arch/power/types.hh | 3 + src/arch/power/utility.cc | 3 +- src/arch/power/utility.hh | 15 + src/arch/sparc/faults.cc | 65 ++- src/arch/sparc/isa/base.isa | 8 + src/arch/sparc/isa/decoder.isa | 89 ++-- src/arch/sparc/isa/formats/branch.isa | 17 +- src/arch/sparc/isa/formats/micro.isa | 16 +- src/arch/sparc/isa/operands.isa | 3 +- src/arch/sparc/nativetrace.cc | 5 +- src/arch/sparc/predecoder.hh | 4 +- src/arch/sparc/process.cc | 25 +- src/arch/sparc/remote_gdb.cc | 21 +- src/arch/sparc/types.hh | 3 + src/arch/sparc/utility.cc | 12 +- src/arch/sparc/utility.hh | 17 + src/arch/x86/faults.cc | 26 +- src/arch/x86/insts/macroop.hh | 3 +- src/arch/x86/insts/microop.hh | 9 + src/arch/x86/insts/static_inst.hh | 6 + src/arch/x86/isa/decoder/two_byte_opcodes.isa | 2 +- src/arch/x86/isa/formats/unknown.isa | 4 +- src/arch/x86/isa/microops/regop.isa | 13 +- src/arch/x86/isa/microops/seqop.isa | 16 +- src/arch/x86/isa/operands.isa | 5 +- src/arch/x86/nativetrace.cc | 2 +- src/arch/x86/predecoder.hh | 28 +- src/arch/x86/process.cc | 10 +- src/arch/x86/system.cc | 3 +- src/arch/x86/tlb.cc | 2 +- src/arch/x86/types.hh | 3 + src/arch/x86/utility.cc | 9 +- src/arch/x86/utility.hh | 16 + src/base/remote_gdb.cc | 16 +- src/base/types.hh | 22 + src/cpu/base_dyn_inst.hh | 130 ++---- src/cpu/base_dyn_inst_impl.hh | 58 +-- src/cpu/checker/cpu.hh | 8 +- src/cpu/exetrace.cc | 10 +- src/cpu/exetrace.hh | 12 +- src/cpu/inorder/comm.hh | 4 +- src/cpu/inorder/cpu.cc | 91 +--- src/cpu/inorder/cpu.hh | 30 +- src/cpu/inorder/first_stage.cc | 4 +- src/cpu/inorder/inorder_dyn_inst.cc | 25 +- src/cpu/inorder/inorder_dyn_inst.hh | 87 +--- src/cpu/inorder/inorder_trace.cc | 8 +- src/cpu/inorder/inorder_trace.hh | 13 +- src/cpu/inorder/pipeline_stage.cc | 49 +- src/cpu/inorder/pipeline_stage.hh | 6 +- src/cpu/inorder/resources/bpred_unit.cc | 94 ++-- src/cpu/inorder/resources/bpred_unit.hh | 44 +- src/cpu/inorder/resources/branch_predictor.cc | 28 +- src/cpu/inorder/resources/cache_unit.cc | 33 +- src/cpu/inorder/resources/execution_unit.cc | 59 ++- src/cpu/inorder/resources/fetch_seq_unit.cc | 171 +++---- src/cpu/inorder/resources/fetch_seq_unit.hh | 17 +- src/cpu/inorder/resources/tlb_unit.hh | 12 +- src/cpu/inorder/thread_context.cc | 24 - src/cpu/inorder/thread_context.hh | 22 +- src/cpu/inorder/thread_state.hh | 4 +- src/cpu/inteltrace.cc | 2 +- src/cpu/inteltrace.hh | 12 +- src/cpu/legiontrace.cc | 9 +- src/cpu/legiontrace.hh | 12 +- src/cpu/nativetrace.hh | 12 +- src/cpu/o3/bpred_unit.hh | 33 +- src/cpu/o3/bpred_unit_impl.hh | 103 ++--- src/cpu/o3/comm.hh | 13 +- src/cpu/o3/commit.hh | 55 +-- src/cpu/o3/commit_impl.hh | 86 ++-- src/cpu/o3/cpu.cc | 83 +--- src/cpu/o3/cpu.hh | 33 +- src/cpu/o3/decode_impl.hh | 51 +-- src/cpu/o3/dep_graph.hh | 8 +- src/cpu/o3/dyn_inst.hh | 12 +- src/cpu/o3/dyn_inst_impl.hh | 32 +- src/cpu/o3/fetch.hh | 20 +- src/cpu/o3/fetch_impl.hh | 178 +++----- src/cpu/o3/iew_impl.hh | 94 ++-- src/cpu/o3/inst_queue_impl.hh | 49 +- src/cpu/o3/lsq_unit.hh | 8 +- src/cpu/o3/lsq_unit_impl.hh | 44 +- src/cpu/o3/mem_dep_unit_impl.hh | 52 +-- src/cpu/o3/rename_impl.hh | 23 +- src/cpu/o3/rob_impl.hh | 8 +- src/cpu/o3/thread_context.hh | 42 +- src/cpu/o3/thread_context_impl.hh | 46 +- src/cpu/pc_event.cc | 4 +- src/cpu/pred/btb.cc | 28 +- src/cpu/pred/btb.hh | 17 +- src/cpu/pred/ras.cc | 15 +- src/cpu/pred/ras.hh | 12 +- src/cpu/simple/atomic.cc | 13 +- src/cpu/simple/base.cc | 82 ++-- src/cpu/simple/base.hh | 23 +- src/cpu/simple/timing.cc | 9 +- src/cpu/simple_thread.cc | 12 +- src/cpu/simple_thread.hh | 82 +--- src/cpu/static_inst.cc | 11 +- src/cpu/static_inst.hh | 34 +- src/cpu/thread_context.cc | 12 +- src/cpu/thread_context.hh | 42 +- src/kern/system_events.cc | 5 +- src/kern/tru64/tru64.hh | 7 +- src/sim/faults.cc | 5 +- src/sim/insttracer.hh | 20 +- src/sim/syscall_emul.cc | 4 +- src/sim/syscall_emul.hh | 4 +- 171 files changed, 2503 insertions(+), 2293 deletions(-) create mode 100644 src/arch/generic/types.hh diff --git a/src/arch/alpha/ev5.cc b/src/arch/alpha/ev5.cc index 0db75df46..f97244260 100644 --- a/src/arch/alpha/ev5.cc +++ b/src/arch/alpha/ev5.cc @@ -60,8 +60,7 @@ initCPU(ThreadContext *tc, int cpuId) AlphaFault *reset = new ResetFault; - tc->setPC(tc->readMiscRegNoEffect(IPR_PAL_BASE) + reset->vect()); - tc->setNextPC(tc->readPC() + sizeof(MachInst)); + tc->pcState(tc->readMiscRegNoEffect(IPR_PAL_BASE) + reset->vect()); delete reset; } @@ -494,12 +493,14 @@ using namespace AlphaISA; Fault SimpleThread::hwrei() { - if (!(readPC() & 0x3)) + PCState pc = pcState(); + if (!(pc.pc() & 0x3)) return new UnimplementedOpcodeFault; - setNextPC(readMiscRegNoEffect(IPR_EXC_ADDR)); + pc.npc(readMiscRegNoEffect(IPR_EXC_ADDR)); + pcState(pc); - CPA::cpa()->swAutoBegin(tc, readNextPC()); + CPA::cpa()->swAutoBegin(tc, pc.npc()); if (!misspeculating()) { if (kernelStats) diff --git a/src/arch/alpha/faults.cc b/src/arch/alpha/faults.cc index 9d4eeda8a..38386cce1 100644 --- a/src/arch/alpha/faults.cc +++ b/src/arch/alpha/faults.cc @@ -115,9 +115,11 @@ AlphaFault::invoke(ThreadContext *tc, StaticInstPtr inst) FaultBase::invoke(tc); countStat()++; + PCState pc = tc->pcState(); + // exception restart address - if (setRestartAddress() || !(tc->readPC() & 0x3)) - tc->setMiscRegNoEffect(IPR_EXC_ADDR, tc->readPC()); + if (setRestartAddress() || !(pc.pc() & 0x3)) + tc->setMiscRegNoEffect(IPR_EXC_ADDR, pc.pc()); if (skipFaultingInstruction()) { // traps... skip faulting instruction. @@ -125,8 +127,8 @@ AlphaFault::invoke(ThreadContext *tc, StaticInstPtr inst) tc->readMiscRegNoEffect(IPR_EXC_ADDR) + 4); } - tc->setPC(tc->readMiscRegNoEffect(IPR_PAL_BASE) + vect()); - tc->setNextPC(tc->readPC() + sizeof(MachInst)); + pc.set(tc->readMiscRegNoEffect(IPR_PAL_BASE) + vect()); + tc->pcState(pc); } void diff --git a/src/arch/alpha/interrupts.hh b/src/arch/alpha/interrupts.hh index 3200201db..cbaa8e9bf 100644 --- a/src/arch/alpha/interrupts.hh +++ b/src/arch/alpha/interrupts.hh @@ -133,7 +133,7 @@ class Interrupts : public SimObject bool checkInterrupts(ThreadContext *tc) const { - return (intstatus != 0) && !(tc->readPC() & 0x3); + return (intstatus != 0) && !(tc->pcState().pc() & 0x3); } Fault diff --git a/src/arch/alpha/isa/branch.isa b/src/arch/alpha/isa/branch.isa index 974193efd..feb15b158 100644 --- a/src/arch/alpha/isa/branch.isa +++ b/src/arch/alpha/isa/branch.isa @@ -81,7 +81,7 @@ output header {{ { } - Addr branchTarget(Addr branchPC) const; + AlphaISA::PCState branchTarget(const AlphaISA::PCState &branchPC) const; std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const; @@ -106,7 +106,7 @@ output header {{ { } - Addr branchTarget(ThreadContext *tc) const; + AlphaISA::PCState branchTarget(ThreadContext *tc) const; std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const; @@ -114,18 +114,19 @@ output header {{ }}; output decoder {{ - Addr - Branch::branchTarget(Addr branchPC) const + AlphaISA::PCState + Branch::branchTarget(const AlphaISA::PCState &branchPC) const { - return branchPC + 4 + disp; + return branchPC.pc() + 4 + disp; } - Addr + AlphaISA::PCState Jump::branchTarget(ThreadContext *tc) const { - Addr NPC = tc->readPC() + 4; + PCState pc = tc->pcState(); uint64_t Rb = tc->readIntReg(_srcRegIdx[0]); - return (Rb & ~3) | (NPC & 1); + pc.set((Rb & ~3) | (pc.pc() & 1)); + return pc; } const std::string & @@ -217,7 +218,14 @@ def template JumpOrBranchDecode {{ }}; def format CondBranch(code) {{ - code = 'bool cond;\n' + code + '\nif (cond) NPC = NPC + disp;\n'; + code = ''' + bool cond; + %(code)s; + PCState pc = PCS; + if (cond) + pc.npc(pc.npc() + disp); + PCS = pc; + ''' % { "code" : code } iop = InstObjParams(name, Name, 'Branch', code, ('IsDirectControl', 'IsCondControl')) header_output = BasicDeclare.subst(iop) @@ -229,16 +237,18 @@ def format CondBranch(code) {{ let {{ def UncondCtrlBase(name, Name, base_class, npc_expr, flags): # Declare basic control transfer w/o link (i.e. link reg is R31) - nolink_code = 'NPC = %s;\n' % npc_expr - nolink_iop = InstObjParams(name, Name, base_class, nolink_code, flags) + readpc_code = 'PCState pc = PCS;' + nolink_code = 'pc.npc(%s);\nPCS = pc' % npc_expr + nolink_iop = InstObjParams(name, Name, base_class, + readpc_code + nolink_code, flags) header_output = BasicDeclare.subst(nolink_iop) decoder_output = BasicConstructor.subst(nolink_iop) exec_output = BasicExecute.subst(nolink_iop) # Generate declaration of '*AndLink' version, append to decls - link_code = 'Ra = NPC & ~3;\n' + nolink_code + link_code = 'Ra = pc.npc() & ~3;\n' + nolink_code link_iop = InstObjParams(name, Name + 'AndLink', base_class, - link_code, flags) + readpc_code + link_code, flags) header_output += BasicDeclare.subst(link_iop) decoder_output += BasicConstructor.subst(link_iop) exec_output += BasicExecute.subst(link_iop) @@ -253,13 +263,13 @@ def UncondCtrlBase(name, Name, base_class, npc_expr, flags): def format UncondBranch(*flags) {{ flags += ('IsUncondControl', 'IsDirectControl') (header_output, decoder_output, decode_block, exec_output) = \ - UncondCtrlBase(name, Name, 'Branch', 'NPC + disp', flags) + UncondCtrlBase(name, Name, 'Branch', 'pc.npc() + disp', flags) }}; def format Jump(*flags) {{ flags += ('IsUncondControl', 'IsIndirectControl') (header_output, decoder_output, decode_block, exec_output) = \ - UncondCtrlBase(name, Name, 'Jump', '(Rb & ~3) | (NPC & 1)', flags) + UncondCtrlBase(name, Name, 'Jump', '(Rb & ~3) | (pc.npc() & 1)', flags) }}; diff --git a/src/arch/alpha/isa/decoder.isa b/src/arch/alpha/isa/decoder.isa index fe70e4d16..e2947cf4a 100644 --- a/src/arch/alpha/isa/decoder.isa +++ b/src/arch/alpha/isa/decoder.isa @@ -856,15 +856,16 @@ decode OPCODE default Unknown::unknown() { // invalid pal function code, or attempt to do privileged // PAL call in non-kernel mode fault = new UnimplementedOpcodeFault; - } - else { + } else { // check to see if simulator wants to do something special // on this PAL call (including maybe suppress it) bool dopal = xc->simPalCheck(palFunc); if (dopal) { - xc->setMiscReg(IPR_EXC_ADDR, NPC); - NPC = xc->readMiscReg(IPR_PAL_BASE) + palOffset; + PCState pc = PCS; + xc->setMiscReg(IPR_EXC_ADDR, pc.npc()); + pc.npc(xc->readMiscReg(IPR_PAL_BASE) + palOffset); + PCS = pc; } } }}, IsNonSpeculative); @@ -1030,13 +1031,14 @@ decode OPCODE default Unknown::unknown() { }}, IsNonSpeculative); #endif 0x54: m5panic({{ - panic("M5 panic instruction called at pc=%#x.", xc->readPC()); + panic("M5 panic instruction called at pc=%#x.", + xc->pcState().pc()); }}, IsNonSpeculative); #define CPANN(lbl) CPA::cpa()->lbl(xc->tcBase()) 0x55: decode RA { 0x00: m5a_old({{ panic("Deprecated M5 annotate instruction executed at pc=%#x\n", - xc->readPC()); + xc->pcState().pc()); }}, IsNonSpeculative); 0x01: m5a_bsm({{ CPANN(swSmBegin); diff --git a/src/arch/alpha/isa/main.isa b/src/arch/alpha/isa/main.isa index 2a0699354..ffc267cd2 100644 --- a/src/arch/alpha/isa/main.isa +++ b/src/arch/alpha/isa/main.isa @@ -46,6 +46,7 @@ output header {{ #include #include "arch/alpha/faults.hh" +#include "arch/alpha/types.hh" #include "config/ss_compatible_fp.hh" #include "cpu/static_inst.hh" #include "mem/request.hh" // some constructors use MemReq flags @@ -185,7 +186,7 @@ def operands {{ 'Fb': ('FloatReg', 'df', 'FB', 'IsFloating', 2), 'Fc': ('FloatReg', 'df', 'FC', 'IsFloating', 3), 'Mem': ('Mem', 'uq', None, ('IsMemRef', 'IsLoad', 'IsStore'), 4), - 'NPC': ('NPC', 'uq', None, ( None, None, 'IsControl' ), 4), + 'PCS': ('PCState', 'uq', None, ( None, None, 'IsControl' ), 4), 'Runiq': ('ControlReg', 'uq', 'MISCREG_UNIQ', None, 1), 'FPCR': ('ControlReg', 'uq', 'MISCREG_FPCR', None, 1), 'IntrFlag': ('ControlReg', 'uq', 'MISCREG_INTR', None, 1), @@ -233,6 +234,12 @@ output header {{ std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const; + + void + advancePC(AlphaISA::PCState &pcState) const + { + pcState.advance(); + } }; }}; diff --git a/src/arch/alpha/predecoder.hh b/src/arch/alpha/predecoder.hh index 913bd8764..f9a716b7f 100644 --- a/src/arch/alpha/predecoder.hh +++ b/src/arch/alpha/predecoder.hh @@ -76,11 +76,11 @@ class Predecoder // Use this to give data to the predecoder. This should be used // when there is control flow. void - moreBytes(Addr pc, Addr fetchPC, MachInst inst) + moreBytes(const PCState &pc, Addr fetchPC, MachInst inst) { ext_inst = inst; #if FULL_SYSTEM - ext_inst |= (static_cast(pc & 0x1) << 32); + ext_inst |= (static_cast(pc.pc() & 0x1) << 32); #endif } @@ -98,7 +98,7 @@ class Predecoder // This returns a constant reference to the ExtMachInst to avoid a copy const ExtMachInst & - getExtMachInst() + getExtMachInst(PCState &pc) { return ext_inst; } diff --git a/src/arch/alpha/process.cc b/src/arch/alpha/process.cc index f68a53a6f..3c3fd9b25 100644 --- a/src/arch/alpha/process.cc +++ b/src/arch/alpha/process.cc @@ -162,15 +162,7 @@ AlphaLiveProcess::argsInit(int intSize, int pageSize) setSyscallArg(tc, 1, argv_array_base); tc->setIntReg(StackPointerReg, stack_min); - Addr prog_entry = objFile->entryPoint(); - tc->setPC(prog_entry); - tc->setNextPC(prog_entry + sizeof(MachInst)); - - // MIPS/Sparc need NNPC for delay slot handling, while - // Alpha has no delay slots... However, CPU models - // cycle PCs by PC=NPC, NPC=NNPC, etc. so setting this - // here ensures CPU-Model Compatibility across board - tc->setNextNPC(prog_entry + (2 * sizeof(MachInst))); + tc->pcState(objFile->entryPoint()); } void diff --git a/src/arch/alpha/remote_gdb.cc b/src/arch/alpha/remote_gdb.cc index 5391d2056..9a2a5f23f 100644 --- a/src/arch/alpha/remote_gdb.cc +++ b/src/arch/alpha/remote_gdb.cc @@ -211,7 +211,7 @@ RemoteGDB::getregs() { memset(gdbregs.regs, 0, gdbregs.bytes()); - gdbregs.regs[KGDB_REG_PC] = context->readPC(); + gdbregs.regs[KGDB_REG_PC] = context->pcState().pc(); // @todo: Currently this is very Alpha specific. if (PcPAL(gdbregs.regs[KGDB_REG_PC])) { @@ -254,7 +254,7 @@ RemoteGDB::setregs() context->setFloatRegBits(i, gdbregs.regs[i + KGDB_REG_F0]); } #endif - context->setPC(gdbregs.regs[KGDB_REG_PC]); + context->pcState(gdbregs.regs[KGDB_REG_PC]); } void @@ -273,30 +273,28 @@ RemoteGDB::clearSingleStep() void RemoteGDB::setSingleStep() { - Addr pc = context->readPC(); - Addr npc, bpc; + PCState pc = context->pcState(); + PCState bpc; bool set_bt = false; - npc = pc + sizeof(MachInst); - // User was stopped at pc, e.g. the instruction at pc was not // executed. - MachInst inst = read(pc); - StaticInstPtr si(inst, pc); + MachInst inst = read(pc.pc()); + StaticInstPtr si(inst, pc.pc()); if (si->hasBranchTarget(pc, context, bpc)) { // Don't bother setting a breakpoint on the taken branch if it // is the same as the next pc - if (bpc != npc) + if (bpc.pc() != pc.npc()) set_bt = true; } DPRINTF(GDBMisc, "setSingleStep bt_addr=%#x nt_addr=%#x\n", takenBkpt, notTakenBkpt); - setTempBreakpoint(notTakenBkpt = npc); + setTempBreakpoint(notTakenBkpt = pc.npc()); if (set_bt) - setTempBreakpoint(takenBkpt = bpc); + setTempBreakpoint(takenBkpt = bpc.pc()); } // Write bytes to kernel address space for debugger. diff --git a/src/arch/alpha/stacktrace.cc b/src/arch/alpha/stacktrace.cc index 1b5a9be34..9c6b3cff0 100644 --- a/src/arch/alpha/stacktrace.cc +++ b/src/arch/alpha/stacktrace.cc @@ -145,7 +145,7 @@ StackTrace::trace(ThreadContext *_tc, bool is_call) bool usermode = (tc->readMiscRegNoEffect(IPR_DTB_CM) & 0x18) != 0; - Addr pc = tc->readNextPC(); + Addr pc = tc->pcState().pc(); bool kernel = sys->kernelStart <= pc && pc <= sys->kernelEnd; if (usermode) { @@ -168,7 +168,7 @@ StackTrace::trace(ThreadContext *_tc, bool is_call) panic("could not find address %#x", pc); stack.push_back(addr); - pc = tc->readPC(); + pc = tc->pcState().pc(); } while (ksp > bottom) { diff --git a/src/arch/alpha/tlb.cc b/src/arch/alpha/tlb.cc index b578741d9..614061ddd 100644 --- a/src/arch/alpha/tlb.cc +++ b/src/arch/alpha/tlb.cc @@ -443,8 +443,6 @@ TLB::translateInst(RequestPtr req, ThreadContext *tc) Fault TLB::translateData(RequestPtr req, ThreadContext *tc, bool write) { - Addr pc = tc->readPC(); - mode_type mode = (mode_type)DTB_CM_CM(tc->readMiscRegNoEffect(IPR_DTB_CM)); @@ -458,7 +456,7 @@ TLB::translateData(RequestPtr req, ThreadContext *tc, bool write) return new DtbAlignmentFault(req->getVaddr(), req->getFlags(), flags); } - if (PcPAL(pc)) { + if (PcPAL(tc->pcState().pc())) { mode = (req->getFlags() & Request::ALTMODE) ? (mode_type)ALT_MODE_AM( tc->readMiscRegNoEffect(IPR_ALT_MODE)) diff --git a/src/arch/alpha/types.hh b/src/arch/alpha/types.hh index 0d285c3b2..06c0168cf 100644 --- a/src/arch/alpha/types.hh +++ b/src/arch/alpha/types.hh @@ -33,12 +33,15 @@ #define __ARCH_ALPHA_TYPES_HH__ #include "base/types.hh" +#include "arch/generic/types.hh" namespace AlphaISA { typedef uint32_t MachInst; typedef uint64_t ExtMachInst; +typedef GenericISA::SimplePCState PCState; + typedef uint64_t LargestRead; enum annotes diff --git a/src/arch/alpha/utility.cc b/src/arch/alpha/utility.cc index 11560259f..2d56ca9b8 100644 --- a/src/arch/alpha/utility.cc +++ b/src/arch/alpha/utility.cc @@ -77,8 +77,7 @@ copyRegs(ThreadContext *src, ThreadContext *dest) copyMiscRegs(src, dest); // Lastly copy PC/NPC - dest->setPC(src->readPC()); - dest->setNextPC(src->readNextPC()); + dest->pcState(src->pcState()); } void @@ -99,9 +98,9 @@ copyMiscRegs(ThreadContext *src, ThreadContext *dest) void skipFunction(ThreadContext *tc) { - Addr newpc = tc->readIntReg(ReturnAddressReg); - tc->setPC(newpc); - tc->setNextPC(tc->readPC() + sizeof(TheISA::MachInst)); + TheISA::PCState newPC = tc->pcState(); + newPC.set(tc->readIntReg(ReturnAddressReg)); + tc->pcState(newPC); } diff --git a/src/arch/alpha/utility.hh b/src/arch/alpha/utility.hh index 6d5b8baf6..fdac8b8d1 100644 --- a/src/arch/alpha/utility.hh +++ b/src/arch/alpha/utility.hh @@ -37,10 +37,19 @@ #include "arch/alpha/registers.hh" #include "base/misc.hh" #include "config/full_system.hh" +#include "cpu/static_inst.hh" #include "cpu/thread_context.hh" namespace AlphaISA { +inline PCState +buildRetPC(const PCState &curPC, const PCState &callPC) +{ + PCState retPC = callPC; + retPC.advance(); + return retPC; +} + uint64_t getArgument(ThreadContext *tc, int &number, uint16_t size, bool fp); inline bool @@ -95,6 +104,13 @@ void copyRegs(ThreadContext *src, ThreadContext *dest); void copyMiscRegs(ThreadContext *src, ThreadContext *dest); void skipFunction(ThreadContext *tc); + +inline void +advancePC(PCState &pc, const StaticInstPtr inst) +{ + pc.advance(); +} + } // namespace AlphaISA #endif // __ARCH_ALPHA_UTILITY_HH__ diff --git a/src/arch/arm/faults.cc b/src/arch/arm/faults.cc index 6e138119c..54cf5cfe5 100644 --- a/src/arch/arm/faults.cc +++ b/src/arch/arm/faults.cc @@ -104,6 +104,7 @@ ArmFault::invoke(ThreadContext *tc, StaticInstPtr inst) CPSR cpsr = tc->readMiscReg(MISCREG_CPSR); CPSR saved_cpsr = tc->readMiscReg(MISCREG_CPSR) | tc->readIntReg(INTREG_CONDCODES); + Addr curPc M5_VAR_USED = tc->pcState().pc(); cpsr.mode = nextMode(); @@ -116,7 +117,7 @@ ArmFault::invoke(ThreadContext *tc, StaticInstPtr inst) cpsr.i = 1; cpsr.e = sctlr.ee; tc->setMiscReg(MISCREG_CPSR, cpsr); - tc->setIntReg(INTREG_LR, tc->readPC() + + tc->setIntReg(INTREG_LR, curPc + (saved_cpsr.t ? thumbPcOffset() : armPcOffset())); switch (nextMode()) { @@ -139,14 +140,15 @@ ArmFault::invoke(ThreadContext *tc, StaticInstPtr inst) panic("unknown Mode\n"); } - Addr pc M5_VAR_USED = tc->readPC(); - Addr newPc = getVector(tc) | (sctlr.te ? PcTBit : 0); + Addr newPc = getVector(tc); DPRINTF(Faults, "Invoking Fault:%s cpsr:%#x PC:%#x lr:%#x newVec: %#x\n", - name(), cpsr, pc, tc->readIntReg(INTREG_LR), newPc); - tc->setPC(newPc); - tc->setNextPC(newPc + cpsr.t ? 2 : 4 ); - tc->setMicroPC(0); - tc->setNextMicroPC(1); + name(), cpsr, curPc, tc->readIntReg(INTREG_LR), newPc); + PCState pc(newPc); + pc.thumb(cpsr.t); + pc.nextThumb(pc.thumb()); + pc.jazelle(cpsr.j); + pc.nextJazelle(pc.jazelle()); + tc->pcState(pc); } void @@ -193,10 +195,10 @@ SupervisorCall::invoke(ThreadContext *tc, StaticInstPtr inst) tc->syscall(callNum); // Advance the PC since that won't happen automatically. - tc->setPC(tc->readNextPC()); - tc->setNextPC(tc->readNextNPC()); - tc->setMicroPC(0); - tc->setNextMicroPC(1); + PCState pc = tc->pcState(); + assert(inst); + inst->advancePC(pc); + tc->pcState(pc); } #endif // FULL_SYSTEM @@ -223,10 +225,10 @@ FlushPipe::invoke(ThreadContext *tc, StaticInstPtr inst) { // Set the PC to the next instruction of the faulting instruction. // Net effect is simply squashing all instructions behind and // start refetching from the next instruction. - tc->setPC(tc->readNextPC()); - tc->setNextPC(tc->readNextNPC()); - tc->setMicroPC(0); - tc->setNextMicroPC(1); + PCState pc = tc->pcState(); + assert(inst); + inst->advancePC(pc); + tc->pcState(pc); } template void AbortFault::invoke(ThreadContext *tc, diff --git a/src/arch/arm/insts/macromem.hh b/src/arch/arm/insts/macromem.hh index c1c8383da..018cb1de5 100644 --- a/src/arch/arm/insts/macromem.hh +++ b/src/arch/arm/insts/macromem.hh @@ -77,6 +77,18 @@ class MicroOp : public PredOp { flags[IsDelayedCommit] = true; } + + void + advancePC(PCState &pcState) const + { + if (flags[IsLastMicroop]) { + pcState.uEnd(); + } else if (flags[IsMicroop]) { + pcState.uAdvance(); + } else { + pcState.advance(); + } + } }; /** diff --git a/src/arch/arm/insts/mem.hh b/src/arch/arm/insts/mem.hh index 1baba5112..2aa4de1d7 100644 --- a/src/arch/arm/insts/mem.hh +++ b/src/arch/arm/insts/mem.hh @@ -63,8 +63,28 @@ class Swap : public PredOp std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const; }; +class MightBeMicro : public PredOp +{ + protected: + MightBeMicro(const char *mnem, ExtMachInst _machInst, OpClass __opClass) + : PredOp(mnem, _machInst, __opClass) + {} + + void + advancePC(PCState &pcState) const + { + if (flags[IsLastMicroop]) { + pcState.uEnd(); + } else if (flags[IsMicroop]) { + pcState.uAdvance(); + } else { + pcState.advance(); + } + } +}; + // The address is a base register plus an immediate. -class RfeOp : public PredOp +class RfeOp : public MightBeMicro { public: enum AddrMode { @@ -83,7 +103,7 @@ class RfeOp : public PredOp RfeOp(const char *mnem, ExtMachInst _machInst, OpClass __opClass, IntRegIndex _base, AddrMode _mode, bool _wb) - : PredOp(mnem, _machInst, __opClass), + : MightBeMicro(mnem, _machInst, __opClass), base(_base), mode(_mode), wb(_wb), uops(NULL) {} @@ -94,7 +114,7 @@ class RfeOp : public PredOp } StaticInstPtr - fetchMicroop(MicroPC microPC) + fetchMicroop(MicroPC microPC) const { assert(uops != NULL && microPC < numMicroops); return uops[microPC]; @@ -104,7 +124,7 @@ class RfeOp : public PredOp }; // The address is a base register plus an immediate. -class SrsOp : public PredOp +class SrsOp : public MightBeMicro { public: enum AddrMode { @@ -123,7 +143,7 @@ class SrsOp : public PredOp SrsOp(const char *mnem, ExtMachInst _machInst, OpClass __opClass, uint32_t _regMode, AddrMode _mode, bool _wb) - : PredOp(mnem, _machInst, __opClass), + : MightBeMicro(mnem, _machInst, __opClass), regMode(_regMode), mode(_mode), wb(_wb), uops(NULL) {} @@ -134,7 +154,7 @@ class SrsOp : public PredOp } StaticInstPtr - fetchMicroop(MicroPC microPC) + fetchMicroop(MicroPC microPC) const { assert(uops != NULL && microPC < numMicroops); return uops[microPC]; @@ -143,7 +163,7 @@ class SrsOp : public PredOp std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const; }; -class Memory : public PredOp +class Memory : public MightBeMicro { public: enum AddrMode { @@ -163,7 +183,7 @@ class Memory : public PredOp Memory(const char *mnem, ExtMachInst _machInst, OpClass __opClass, IntRegIndex _dest, IntRegIndex _base, bool _add) - : PredOp(mnem, _machInst, __opClass), + : MightBeMicro(mnem, _machInst, __opClass), dest(_dest), base(_base), add(_add), uops(NULL) {} @@ -174,7 +194,7 @@ class Memory : public PredOp } StaticInstPtr - fetchMicroop(MicroPC microPC) + fetchMicroop(MicroPC microPC) const { assert(uops != NULL && microPC < numMicroops); return uops[microPC]; diff --git a/src/arch/arm/insts/pred_inst.hh b/src/arch/arm/insts/pred_inst.hh index b7d4c4709..f779b46f5 100644 --- a/src/arch/arm/insts/pred_inst.hh +++ b/src/arch/arm/insts/pred_inst.hh @@ -312,7 +312,7 @@ class PredMacroOp : public PredOp } StaticInstPtr - fetchMicroop(MicroPC microPC) + fetchMicroop(MicroPC microPC) const { assert(microPC < numMicroops); return microOps[microPC]; @@ -332,6 +332,15 @@ class PredMicroop : public PredOp { flags[IsMicroop] = true; } + + void + advancePC(PCState &pcState) const + { + if (flags[IsLastMicroop]) + pcState.uEnd(); + else + pcState.uAdvance(); + } }; } diff --git a/src/arch/arm/insts/static_inst.hh b/src/arch/arm/insts/static_inst.hh index ad89a1a79..fa850190f 100644 --- a/src/arch/arm/insts/static_inst.hh +++ b/src/arch/arm/insts/static_inst.hh @@ -155,6 +155,12 @@ class ArmStaticInst : public StaticInst IntRegIndex rs, uint32_t shiftAmt, ArmShiftType type, uint32_t imm) const; + void + advancePC(PCState &pcState) const + { + pcState.advance(); + } + std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const; static inline uint32_t @@ -219,26 +225,16 @@ class ArmStaticInst : public StaticInst static inline Addr readPC(XC *xc) { - Addr pc = xc->readPC(); - if (isThumb(pc)) - return pc + 4; - else - return pc + 8; + return xc->pcState().instPC(); } - // Perform an regular branch. template static inline void setNextPC(XC *xc, Addr val) { - Addr npc = xc->readNextPC(); - if (isThumb(npc)) { - val &= ~mask(1); - } else { - val &= ~mask(2); - } - xc->setNextPC((npc & PcModeMask) | - (val & ~PcModeMask)); + PCState pc = xc->pcState(); + pc.instNPC(val); + xc->pcState(pc); } template @@ -279,30 +275,9 @@ class ArmStaticInst : public StaticInst static inline void setIWNextPC(XC *xc, Addr val) { - Addr stateBits = xc->readPC() & PcModeMask; - Addr jBit = PcJBit; - Addr tBit = PcTBit; - bool thumbEE = (stateBits == (tBit | jBit)); - - Addr newPc = (val & ~PcModeMask); - if (thumbEE) { - if (bits(newPc, 0)) { - newPc = newPc & ~mask(1); - } else { - panic("Bad thumbEE interworking branch address %#x.\n", newPc); - } - } else { - if (bits(newPc, 0)) { - stateBits = tBit; - newPc = newPc & ~mask(1); - } else if (!bits(newPc, 1)) { - stateBits = 0; - } else { - warn("Bad interworking branch address %#x.\n", newPc); - } - } - newPc = newPc | stateBits; - xc->setNextPC(newPc); + PCState pc = xc->pcState(); + pc.instIWNPC(val); + xc->pcState(pc); } // Perform an interworking branch in ARM mode, a regular branch @@ -311,14 +286,9 @@ class ArmStaticInst : public StaticInst static inline void setAIWNextPC(XC *xc, Addr val) { - Addr stateBits = xc->readPC() & PcModeMask; - Addr jBit = PcJBit; - Addr tBit = PcTBit; - if (!jBit && !tBit) { - setIWNextPC(xc, val); - } else { - setNextPC(xc, val); - } + PCState pc = xc->pcState(); + pc.instAIWNPC(val); + xc->pcState(pc); } inline Fault diff --git a/src/arch/arm/insts/vfp.hh b/src/arch/arm/insts/vfp.hh index 964b62673..e962704e0 100644 --- a/src/arch/arm/insts/vfp.hh +++ b/src/arch/arm/insts/vfp.hh @@ -459,6 +459,18 @@ class FpOp : public PredOp unaryOp(FPSCR &fpscr, fpType op1, fpType (*func)(fpType), bool flush, uint32_t rMode) const; + + void + advancePC(PCState &pcState) const + { + if (flags[IsLastMicroop]) { + pcState.uEnd(); + } else if (flags[IsMicroop]) { + pcState.uAdvance(); + } else { + pcState.advance(); + } + } }; class FpRegRegOp : public FpOp diff --git a/src/arch/arm/isa.cc b/src/arch/arm/isa.cc index d557cecbb..0ba62f08d 100644 --- a/src/arch/arm/isa.cc +++ b/src/arch/arm/isa.cc @@ -168,15 +168,9 @@ ISA::readMiscReg(int misc_reg, ThreadContext *tc) { if (misc_reg == MISCREG_CPSR) { CPSR cpsr = miscRegs[misc_reg]; - Addr pc = tc->readPC(); - if (pc & (ULL(1) << PcJBitShift)) - cpsr.j = 1; - else - cpsr.j = 0; - if (isThumb(pc)) - cpsr.t = 1; - else - cpsr.t = 0; + PCState pc = tc->pcState(); + cpsr.j = pc.jazelle() ? 1 : 0; + cpsr.t = pc.thumb() ? 1 : 0; return cpsr; } if (misc_reg >= MISCREG_CP15_UNIMP_START && @@ -239,13 +233,10 @@ ISA::setMiscReg(int misc_reg, const MiscReg &val, ThreadContext *tc) CPSR cpsr = val; DPRINTF(Arm, "Updating CPSR from %#x to %#x f:%d i:%d a:%d mode:%#x\n", miscRegs[misc_reg], cpsr, cpsr.f, cpsr.i, cpsr.a, cpsr.mode); - Addr npc = tc->readNextPC() & ~PcModeMask; - if (cpsr.j) - npc = npc | PcJBit; - if (cpsr.t) - npc = npc | PcTBit; - - tc->setNextPC(npc); + PCState pc = tc->pcState(); + pc.nextThumb(cpsr.t); + pc.nextJazelle(cpsr.j); + tc->pcState(pc); } else if (misc_reg >= MISCREG_CP15_UNIMP_START && misc_reg < MISCREG_CP15_END) { panic("Unimplemented CP15 register %s wrote with %#x.\n", @@ -414,7 +405,7 @@ ISA::setMiscReg(int misc_reg, const MiscReg &val, ThreadContext *tc) default: panic("Security Extensions not implemented!"); } - req->setVirt(0, val, 1, flags, tc->readPC()); + req->setVirt(0, val, 1, flags, tc->pcState().pc()); fault = tc->getDTBPtr()->translateAtomic(req, tc, mode); if (fault == NoFault) { miscRegs[MISCREG_PAR] = diff --git a/src/arch/arm/isa/formats/breakpoint.isa b/src/arch/arm/isa/formats/breakpoint.isa index d59f6a712..1825d0878 100644 --- a/src/arch/arm/isa/formats/breakpoint.isa +++ b/src/arch/arm/isa/formats/breakpoint.isa @@ -83,7 +83,7 @@ output exec {{ Breakpoint::execute(%(CPU_exec_context)s *xc, Trace::InstRecord *traceData) const { - return new PrefetchAbort(xc->readPC(), ArmFault::DebugEvent); + return new PrefetchAbort(xc->pcState().pc(), ArmFault::DebugEvent); } }}; diff --git a/src/arch/arm/isa/insts/branch.isa b/src/arch/arm/isa/insts/branch.isa index e9ddd77b7..3ff9042e6 100644 --- a/src/arch/arm/isa/insts/branch.isa +++ b/src/arch/arm/isa/insts/branch.isa @@ -46,15 +46,17 @@ let {{ # B, BL for (mnem, link) in (("b", False), ("bl", True)): bCode = ''' - Addr curPc = readPC(xc); - NPC = ((curPc + imm) & mask(32)) | (curPc & ~mask(32)); + ArmISA::PCState pc = PCS; + Addr curPc = pc.instPC(); + pc.instNPC((uint32_t)(curPc + imm)); + PCS = pc; ''' if (link): bCode += ''' - if (!isThumb(curPc)) - LR = curPc - 4; - else + if (pc.thumb()) LR = curPc | 1; + else + LR = curPc - 4; ''' bIop = InstObjParams(mnem, mnem.capitalize(), "BranchImmCond", @@ -66,10 +68,12 @@ let {{ # BX, BLX blxCode = ''' - Addr curPc M5_VAR_USED = readPC(xc); + ArmISA::PCState pc = PCS; + Addr curPc M5_VAR_USED = pc.instPC(); %(link)s // Switch modes %(branch)s + PCS = pc; ''' blxList = (("blx", True, True), @@ -81,8 +85,8 @@ let {{ if imm: Name += "Imm" # Since we're switching ISAs, the target ISA will be the opposite - # of the current ISA. !arm is whether the target is ARM. - newPC = '(isThumb(curPc) ? (roundDown(curPc, 4) + imm) : (curPc + imm))' + # of the current ISA. pc.thumb() is whether the target is ARM. + newPC = '(pc.thumb() ? (roundDown(curPc, 4) + imm) : (curPc + imm))' base = "BranchImmCond" declare = BranchImmCondDeclare constructor = BranchImmCondConstructor @@ -97,28 +101,28 @@ let {{ // The immediate version of the blx thumb instruction // is 32 bits wide, but "next pc" doesn't reflect that // so we don't want to substract 2 from it at this point - if (!isThumb(curPc)) - LR = curPc - 4; - else + if (pc.thumb()) LR = curPc | 1; + else + LR = curPc - 4; ''' elif link: linkStr = ''' - if (!isThumb(curPc)) - LR = curPc - 4; - else + if (pc.thumb()) LR = (curPc - 2) | 1; + else + LR = curPc - 4; ''' else: linkStr = "" if imm and link: #blx with imm branchStr = ''' - Addr tempPc = ((%(newPC)s) & mask(32)) | (curPc & ~mask(32)); - FNPC = tempPc ^ PcTBit; + pc.nextThumb(!pc.thumb()); + pc.instNPC(%(newPC)s); ''' else: - branchStr = "IWNPC = %(newPC)s;" + branchStr = "pc.instIWNPC(%(newPC)s);" branchStr = branchStr % { "newPC" : newPC } code = blxCode % {"link": linkStr, @@ -136,8 +140,10 @@ let {{ #CBNZ, CBZ. These are always unconditional as far as predicates for (mnem, test) in (("cbz", "=="), ("cbnz", "!=")): code = ''' - Addr curPc = readPC(xc); - NPC = ((curPc + imm) & mask(32)) | (curPc & ~mask(32)); + ArmISA::PCState pc = PCS; + Addr curPc = pc.instPC(); + pc.instNPC((uint32_t)(curPc + imm)); + PCS = pc; ''' predTest = "Op1 %(test)s 0" % {"test": test} iop = InstObjParams(mnem, mnem.capitalize(), "BranchImmReg", @@ -155,7 +161,11 @@ let {{ ArmISA::TLB::MustBeOne; EA = Op1 + Op2 * 2 ''' - accCode = "NPC = readPC(xc) + 2 * (Mem.uh);" + accCode = ''' + ArmISA::PCState pc = PCS; + pc.instNPC(pc.instPC() + 2 * (Mem.uh)); + PCS = pc; + ''' mnem = "tbh" else: eaCode = ''' @@ -164,7 +174,11 @@ let {{ ArmISA::TLB::MustBeOne; EA = Op1 + Op2 ''' - accCode = "NPC = readPC(xc) + 2 * (Mem.ub);" + accCode = ''' + ArmISA::PCState pc = PCS; + pc.instNPC(pc.instPC() + 2 * (Mem.ub)); + PCS = pc; + ''' mnem = "tbb" iop = InstObjParams(mnem, mnem.capitalize(), "BranchRegReg", {'ea_code': eaCode, diff --git a/src/arch/arm/isa/insts/data.isa b/src/arch/arm/isa/insts/data.isa index 74eeee3b2..4d368e181 100644 --- a/src/arch/arm/isa/insts/data.isa +++ b/src/arch/arm/isa/insts/data.isa @@ -239,6 +239,10 @@ let {{ cpsrWriteByInstr(Cpsr | CondCodes, Spsr, 0xF, true, sctlr.nmfi); Cpsr = ~CondCodesMask & newCpsr; CondCodes = CondCodesMask & newCpsr; + ArmISA::PCState pc = PCS; + pc.nextThumb(((CPSR)newCpsr).t); + pc.nextJazelle(((CPSR)newCpsr).j); + PCS = pc; ''' buildImmDataInst(mnem + 's', code, flagType, suffix = "ImmPclr", buildCc = False, @@ -253,7 +257,8 @@ let {{ buildDataInst("rsb", "Dest = resTemp = secondOp - Op1;", "rsb") buildDataInst("add", "Dest = resTemp = Op1 + secondOp;", "add") buildImmDataInst("adr", ''' - Dest = resTemp = (readPC(xc) & ~0x3) + + ArmISA::PCState pc = PCS; + Dest = resTemp = (pc.instPC() & ~0x3) + (op1 ? secondOp : -secondOp); ''') buildDataInst("adc", "Dest = resTemp = Op1 + secondOp + %s;" % oldC, "add") diff --git a/src/arch/arm/isa/insts/ldr.isa b/src/arch/arm/isa/insts/ldr.isa index dc043ed8e..92ad52a6d 100644 --- a/src/arch/arm/isa/insts/ldr.isa +++ b/src/arch/arm/isa/insts/ldr.isa @@ -105,12 +105,16 @@ let {{ accCode = ''' CPSR cpsr = Cpsr; SCTLR sctlr = Sctlr; - NPC = cSwap(Mem.ud, cpsr.e); + ArmISA::PCState pc = PCS; + pc.instNPC(cSwap(Mem.ud, cpsr.e)); uint32_t newCpsr = cpsrWriteByInstr(cpsr | CondCodes, cSwap(Mem.ud >> 32, cpsr.e), 0xF, true, sctlr.nmfi); Cpsr = ~CondCodesMask & newCpsr; + pc.nextThumb(((CPSR)newCpsr).t); + pc.nextJazelle(((CPSR)newCpsr).j); + PCS = pc; CondCodes = CondCodesMask & newCpsr; ''' self.codeBlobs["memacc_code"] = accCode diff --git a/src/arch/arm/isa/insts/macromem.isa b/src/arch/arm/isa/insts/macromem.isa index 6bf789efd..a81050b1e 100644 --- a/src/arch/arm/isa/insts/macromem.isa +++ b/src/arch/arm/isa/insts/macromem.isa @@ -93,7 +93,9 @@ let {{ cpsrWriteByInstr(cpsr | CondCodes, Spsr, 0xF, true, sctlr.nmfi); Cpsr = ~CondCodesMask & newCpsr; CondCodes = CondCodesMask & newCpsr; - IWNPC = cSwap(Mem.uw, cpsr.e) | ((Spsr & 0x20) ? 1 : 0); + ArmISA::PCState pc = PCS; + pc.instIWNPC(cSwap(Mem.uw, cpsr.e) | ((Spsr & 0x20) ? 1 : 0)); + PCS = pc; ''' microLdrRetUopIop = InstObjParams('ldr_ret_uop', 'MicroLdrRetUop', 'MicroMemOp', diff --git a/src/arch/arm/isa/insts/misc.isa b/src/arch/arm/isa/insts/misc.isa index 5742f84ab..1abbc3de1 100644 --- a/src/arch/arm/isa/insts/misc.isa +++ b/src/arch/arm/isa/insts/misc.isa @@ -83,6 +83,10 @@ let {{ uint32_t newCpsr = cpsrWriteByInstr(Cpsr | CondCodes, Op1, byteMask, false, sctlr.nmfi); Cpsr = ~CondCodesMask & newCpsr; + ArmISA::PCState pc = PCS; + pc.nextThumb(((CPSR)newCpsr).t); + pc.nextJazelle(((CPSR)newCpsr).j); + PCS = pc; CondCodes = CondCodesMask & newCpsr; ''' msrCpsrRegIop = InstObjParams("msr", "MsrCpsrReg", "MsrRegOp", @@ -107,6 +111,10 @@ let {{ uint32_t newCpsr = cpsrWriteByInstr(Cpsr | CondCodes, imm, byteMask, false, sctlr.nmfi); Cpsr = ~CondCodesMask & newCpsr; + ArmISA::PCState pc = PCS; + pc.nextThumb(((CPSR)newCpsr).t); + pc.nextJazelle(((CPSR)newCpsr).j); + PCS = pc; CondCodes = CondCodesMask & newCpsr; ''' msrCpsrImmIop = InstObjParams("msr", "MsrCpsrImm", "MsrImmOp", @@ -462,8 +470,12 @@ let {{ decoder_output += RegRegRegRegOpConstructor.subst(usada8Iop) exec_output += PredOpExecute.subst(usada8Iop) + bkptCode = ''' + ArmISA::PCState pc = PCS; + return new PrefetchAbort(pc.pc(), ArmFault::DebugEvent); + ''' bkptIop = InstObjParams("bkpt", "BkptInst", "ArmStaticInst", - "return new PrefetchAbort(PC, ArmFault::DebugEvent);") + bkptCode) header_output += BasicDeclare.subst(bkptIop) decoder_output += BasicConstructor.subst(bkptIop) exec_output += BasicExecute.subst(bkptIop) @@ -638,7 +650,10 @@ let {{ exec_output += PredOpExecute.subst(mcr15UserIop) enterxCode = ''' - FNPC = NPC | PcJBit | PcTBit; + ArmISA::PCState pc = PCS; + pc.nextThumb(true); + pc.nextJazelle(true); + PCS = pc; ''' enterxIop = InstObjParams("enterx", "Enterx", "PredOp", { "code": enterxCode, @@ -648,7 +663,10 @@ let {{ exec_output += PredOpExecute.subst(enterxIop) leavexCode = ''' - FNPC = (NPC & ~PcJBit) | PcTBit; + ArmISA::PCState pc = PCS; + pc.nextThumb(true); + pc.nextJazelle(false); + PCS = pc; ''' leavexIop = InstObjParams("leavex", "Leavex", "PredOp", { "code": leavexCode, diff --git a/src/arch/arm/isa/operands.isa b/src/arch/arm/isa/operands.isa index 4f1b7f610..8e856e74d 100644 --- a/src/arch/arm/isa/operands.isa +++ b/src/arch/arm/isa/operands.isa @@ -54,11 +54,10 @@ def operand_types {{ let {{ maybePCRead = ''' - ((%(reg_idx)s == PCReg) ? (readPC(xc) & ~PcModeMask) : - xc->%(func)s(this, %(op_idx)s)) + ((%(reg_idx)s == PCReg) ? readPC(xc) : xc->%(func)s(this, %(op_idx)s)) ''' maybeAlignedPCRead = ''' - ((%(reg_idx)s == PCReg) ? (roundDown(readPC(xc) & ~PcModeMask, 4)) : + ((%(reg_idx)s == PCReg) ? (roundDown(readPC(xc), 4)) : xc->%(func)s(this, %(op_idx)s)) ''' maybePCWrite = ''' @@ -81,140 +80,133 @@ let {{ xc->%(func)s(this, %(op_idx)s, %(final_val)s); } ''' - - readNPC = 'xc->readNextPC() & ~PcModeMask' - writeNPC = 'setNextPC(xc, %(final_val)s)' - writeIWNPC = 'setIWNextPC(xc, %(final_val)s)' - forceNPC = 'xc->setNextPC(%(final_val)s)' }}; def operands {{ #Abstracted integer reg operands - 'Dest': ('IntReg', 'uw', 'dest', 'IsInteger', 2, + 'Dest': ('IntReg', 'uw', 'dest', 'IsInteger', 3, maybePCRead, maybePCWrite), - 'FpDest': ('FloatReg', 'sf', '(dest + 0)', 'IsFloating', 2), - 'FpDestP0': ('FloatReg', 'sf', '(dest + 0)', 'IsFloating', 2), - 'FpDestP1': ('FloatReg', 'sf', '(dest + 1)', 'IsFloating', 2), - 'FpDestP2': ('FloatReg', 'sf', '(dest + 2)', 'IsFloating', 2), - 'FpDestP3': ('FloatReg', 'sf', '(dest + 3)', 'IsFloating', 2), - 'FpDestP4': ('FloatReg', 'sf', '(dest + 4)', 'IsFloating', 2), - 'FpDestP5': ('FloatReg', 'sf', '(dest + 5)', 'IsFloating', 2), - 'FpDestP6': ('FloatReg', 'sf', '(dest + 6)', 'IsFloating', 2), - 'FpDestP7': ('FloatReg', 'sf', '(dest + 7)', 'IsFloating', 2), - 'FpDestS0P0': ('FloatReg', 'sf', '(dest + step * 0 + 0)', 'IsFloating', 2), - 'FpDestS0P1': ('FloatReg', 'sf', '(dest + step * 0 + 1)', 'IsFloating', 2), - 'FpDestS1P0': ('FloatReg', 'sf', '(dest + step * 1 + 0)', 'IsFloating', 2), - 'FpDestS1P1': ('FloatReg', 'sf', '(dest + step * 1 + 1)', 'IsFloating', 2), - 'FpDestS2P0': ('FloatReg', 'sf', '(dest + step * 2 + 0)', 'IsFloating', 2), - 'FpDestS2P1': ('FloatReg', 'sf', '(dest + step * 2 + 1)', 'IsFloating', 2), - 'FpDestS3P0': ('FloatReg', 'sf', '(dest + step * 3 + 0)', 'IsFloating', 2), - 'FpDestS3P1': ('FloatReg', 'sf', '(dest + step * 3 + 1)', 'IsFloating', 2), - 'Result': ('IntReg', 'uw', 'result', 'IsInteger', 2, + 'FpDest': ('FloatReg', 'sf', '(dest + 0)', 'IsFloating', 3), + 'FpDestP0': ('FloatReg', 'sf', '(dest + 0)', 'IsFloating', 3), + 'FpDestP1': ('FloatReg', 'sf', '(dest + 1)', 'IsFloating', 3), + 'FpDestP2': ('FloatReg', 'sf', '(dest + 2)', 'IsFloating', 3), + 'FpDestP3': ('FloatReg', 'sf', '(dest + 3)', 'IsFloating', 3), + 'FpDestP4': ('FloatReg', 'sf', '(dest + 4)', 'IsFloating', 3), + 'FpDestP5': ('FloatReg', 'sf', '(dest + 5)', 'IsFloating', 3), + 'FpDestP6': ('FloatReg', 'sf', '(dest + 6)', 'IsFloating', 3), + 'FpDestP7': ('FloatReg', 'sf', '(dest + 7)', 'IsFloating', 3), + 'FpDestS0P0': ('FloatReg', 'sf', '(dest + step * 0 + 0)', 'IsFloating', 3), + 'FpDestS0P1': ('FloatReg', 'sf', '(dest + step * 0 + 1)', 'IsFloating', 3), + 'FpDestS1P0': ('FloatReg', 'sf', '(dest + step * 1 + 0)', 'IsFloating', 3), + 'FpDestS1P1': ('FloatReg', 'sf', '(dest + step * 1 + 1)', 'IsFloating', 3), + 'FpDestS2P0': ('FloatReg', 'sf', '(dest + step * 2 + 0)', 'IsFloating', 3), + 'FpDestS2P1': ('FloatReg', 'sf', '(dest + step * 2 + 1)', 'IsFloating', 3), + 'FpDestS3P0': ('FloatReg', 'sf', '(dest + step * 3 + 0)', 'IsFloating', 3), + 'FpDestS3P1': ('FloatReg', 'sf', '(dest + step * 3 + 1)', 'IsFloating', 3), + 'Result': ('IntReg', 'uw', 'result', 'IsInteger', 3, maybePCRead, maybePCWrite), - 'Dest2': ('IntReg', 'uw', 'dest2', 'IsInteger', 2, + 'Dest2': ('IntReg', 'uw', 'dest2', 'IsInteger', 3, maybePCRead, maybePCWrite), - 'FpDest2': ('FloatReg', 'sf', '(dest2 + 0)', 'IsFloating', 2), - 'FpDest2P0': ('FloatReg', 'sf', '(dest2 + 0)', 'IsFloating', 2), - 'FpDest2P1': ('FloatReg', 'sf', '(dest2 + 1)', 'IsFloating', 2), - 'FpDest2P2': ('FloatReg', 'sf', '(dest2 + 2)', 'IsFloating', 2), - 'FpDest2P3': ('FloatReg', 'sf', '(dest2 + 3)', 'IsFloating', 2), - 'IWDest': ('IntReg', 'uw', 'dest', 'IsInteger', 2, + 'FpDest2': ('FloatReg', 'sf', '(dest2 + 0)', 'IsFloating', 3), + 'FpDest2P0': ('FloatReg', 'sf', '(dest2 + 0)', 'IsFloating', 3), + 'FpDest2P1': ('FloatReg', 'sf', '(dest2 + 1)', 'IsFloating', 3), + 'FpDest2P2': ('FloatReg', 'sf', '(dest2 + 2)', 'IsFloating', 3), + 'FpDest2P3': ('FloatReg', 'sf', '(dest2 + 3)', 'IsFloating', 3), + 'IWDest': ('IntReg', 'uw', 'dest', 'IsInteger', 3, maybePCRead, maybeIWPCWrite), - 'AIWDest': ('IntReg', 'uw', 'dest', 'IsInteger', 2, + 'AIWDest': ('IntReg', 'uw', 'dest', 'IsInteger', 3, maybePCRead, maybeAIWPCWrite), 'SpMode': ('IntReg', 'uw', 'intRegInMode((OperatingMode)regMode, INTREG_SP)', - 'IsInteger', 2), - 'MiscDest': ('ControlReg', 'uw', 'dest', (None, None, 'IsControl'), 2), - 'Base': ('IntReg', 'uw', 'base', 'IsInteger', 0, + 'IsInteger', 3), + 'MiscDest': ('ControlReg', 'uw', 'dest', (None, None, 'IsControl'), 3), + 'Base': ('IntReg', 'uw', 'base', 'IsInteger', 1, maybeAlignedPCRead, maybePCWrite), - 'Index': ('IntReg', 'uw', 'index', 'IsInteger', 2, + 'Index': ('IntReg', 'uw', 'index', 'IsInteger', 3, maybePCRead, maybePCWrite), - 'Op1': ('IntReg', 'uw', 'op1', 'IsInteger', 2, + 'Op1': ('IntReg', 'uw', 'op1', 'IsInteger', 3, maybePCRead, maybePCWrite), - 'FpOp1': ('FloatReg', 'sf', '(op1 + 0)', 'IsFloating', 2), - 'FpOp1P0': ('FloatReg', 'sf', '(op1 + 0)', 'IsFloating', 2), - 'FpOp1P1': ('FloatReg', 'sf', '(op1 + 1)', 'IsFloating', 2), - 'FpOp1P2': ('FloatReg', 'sf', '(op1 + 2)', 'IsFloating', 2), - 'FpOp1P3': ('FloatReg', 'sf', '(op1 + 3)', 'IsFloating', 2), - 'FpOp1P4': ('FloatReg', 'sf', '(op1 + 4)', 'IsFloating', 2), - 'FpOp1P5': ('FloatReg', 'sf', '(op1 + 5)', 'IsFloating', 2), - 'FpOp1P6': ('FloatReg', 'sf', '(op1 + 6)', 'IsFloating', 2), - 'FpOp1P7': ('FloatReg', 'sf', '(op1 + 7)', 'IsFloating', 2), - 'FpOp1S0P0': ('FloatReg', 'sf', '(op1 + step * 0 + 0)', 'IsFloating', 2), - 'FpOp1S0P1': ('FloatReg', 'sf', '(op1 + step * 0 + 1)', 'IsFloating', 2), - 'FpOp1S1P0': ('FloatReg', 'sf', '(op1 + step * 1 + 0)', 'IsFloating', 2), - 'FpOp1S1P1': ('FloatReg', 'sf', '(op1 + step * 1 + 1)', 'IsFloating', 2), - 'FpOp1S2P0': ('FloatReg', 'sf', '(op1 + step * 2 + 0)', 'IsFloating', 2), - 'FpOp1S2P1': ('FloatReg', 'sf', '(op1 + step * 2 + 1)', 'IsFloating', 2), - 'FpOp1S3P0': ('FloatReg', 'sf', '(op1 + step * 3 + 0)', 'IsFloating', 2), - 'FpOp1S3P1': ('FloatReg', 'sf', '(op1 + step * 3 + 1)', 'IsFloating', 2), - 'MiscOp1': ('ControlReg', 'uw', 'op1', (None, None, 'IsControl'), 2), - 'Op2': ('IntReg', 'uw', 'op2', 'IsInteger', 2, + 'FpOp1': ('FloatReg', 'sf', '(op1 + 0)', 'IsFloating', 3), + 'FpOp1P0': ('FloatReg', 'sf', '(op1 + 0)', 'IsFloating', 3), + 'FpOp1P1': ('FloatReg', 'sf', '(op1 + 1)', 'IsFloating', 3), + 'FpOp1P2': ('FloatReg', 'sf', '(op1 + 2)', 'IsFloating', 3), + 'FpOp1P3': ('FloatReg', 'sf', '(op1 + 3)', 'IsFloating', 3), + 'FpOp1P4': ('FloatReg', 'sf', '(op1 + 4)', 'IsFloating', 3), + 'FpOp1P5': ('FloatReg', 'sf', '(op1 + 5)', 'IsFloating', 3), + 'FpOp1P6': ('FloatReg', 'sf', '(op1 + 6)', 'IsFloating', 3), + 'FpOp1P7': ('FloatReg', 'sf', '(op1 + 7)', 'IsFloating', 3), + 'FpOp1S0P0': ('FloatReg', 'sf', '(op1 + step * 0 + 0)', 'IsFloating', 3), + 'FpOp1S0P1': ('FloatReg', 'sf', '(op1 + step * 0 + 1)', 'IsFloating', 3), + 'FpOp1S1P0': ('FloatReg', 'sf', '(op1 + step * 1 + 0)', 'IsFloating', 3), + 'FpOp1S1P1': ('FloatReg', 'sf', '(op1 + step * 1 + 1)', 'IsFloating', 3), + 'FpOp1S2P0': ('FloatReg', 'sf', '(op1 + step * 2 + 0)', 'IsFloating', 3), + 'FpOp1S2P1': ('FloatReg', 'sf', '(op1 + step * 2 + 1)', 'IsFloating', 3), + 'FpOp1S3P0': ('FloatReg', 'sf', '(op1 + step * 3 + 0)', 'IsFloating', 3), + 'FpOp1S3P1': ('FloatReg', 'sf', '(op1 + step * 3 + 1)', 'IsFloating', 3), + 'MiscOp1': ('ControlReg', 'uw', 'op1', (None, None, 'IsControl'), 3), + 'Op2': ('IntReg', 'uw', 'op2', 'IsInteger', 3, maybePCRead, maybePCWrite), - 'FpOp2': ('FloatReg', 'sf', '(op2 + 0)', 'IsFloating', 2), - 'FpOp2P0': ('FloatReg', 'sf', '(op2 + 0)', 'IsFloating', 2), - 'FpOp2P1': ('FloatReg', 'sf', '(op2 + 1)', 'IsFloating', 2), - 'FpOp2P2': ('FloatReg', 'sf', '(op2 + 2)', 'IsFloating', 2), - 'FpOp2P3': ('FloatReg', 'sf', '(op2 + 3)', 'IsFloating', 2), - 'Op3': ('IntReg', 'uw', 'op3', 'IsInteger', 2, + 'FpOp2': ('FloatReg', 'sf', '(op2 + 0)', 'IsFloating', 3), + 'FpOp2P0': ('FloatReg', 'sf', '(op2 + 0)', 'IsFloating', 3), + 'FpOp2P1': ('FloatReg', 'sf', '(op2 + 1)', 'IsFloating', 3), + 'FpOp2P2': ('FloatReg', 'sf', '(op2 + 2)', 'IsFloating', 3), + 'FpOp2P3': ('FloatReg', 'sf', '(op2 + 3)', 'IsFloating', 3), + 'Op3': ('IntReg', 'uw', 'op3', 'IsInteger', 3, maybePCRead, maybePCWrite), - 'Shift': ('IntReg', 'uw', 'shift', 'IsInteger', 2, + 'Shift': ('IntReg', 'uw', 'shift', 'IsInteger', 3, maybePCRead, maybePCWrite), - 'Reg0': ('IntReg', 'uw', 'reg0', 'IsInteger', 2, + 'Reg0': ('IntReg', 'uw', 'reg0', 'IsInteger', 3, maybePCRead, maybePCWrite), - 'Reg1': ('IntReg', 'uw', 'reg1', 'IsInteger', 2, + 'Reg1': ('IntReg', 'uw', 'reg1', 'IsInteger', 3, maybePCRead, maybePCWrite), - 'Reg2': ('IntReg', 'uw', 'reg2', 'IsInteger', 2, + 'Reg2': ('IntReg', 'uw', 'reg2', 'IsInteger', 3, maybePCRead, maybePCWrite), - 'Reg3': ('IntReg', 'uw', 'reg3', 'IsInteger', 2, + 'Reg3': ('IntReg', 'uw', 'reg3', 'IsInteger', 3, maybePCRead, maybePCWrite), #General Purpose Integer Reg Operands - 'Rd': ('IntReg', 'uw', 'RD', 'IsInteger', 2, maybePCRead, maybePCWrite), - 'Rm': ('IntReg', 'uw', 'RM', 'IsInteger', 2, maybePCRead, maybePCWrite), - 'Rs': ('IntReg', 'uw', 'RS', 'IsInteger', 2, maybePCRead, maybePCWrite), - 'Rn': ('IntReg', 'uw', 'RN', 'IsInteger', 2, maybePCRead, maybePCWrite), - 'R7': ('IntReg', 'uw', '7', 'IsInteger', 2), - 'R0': ('IntReg', 'uw', '0', 'IsInteger', 2), + 'Rd': ('IntReg', 'uw', 'RD', 'IsInteger', 3, maybePCRead, maybePCWrite), + 'Rm': ('IntReg', 'uw', 'RM', 'IsInteger', 3, maybePCRead, maybePCWrite), + 'Rs': ('IntReg', 'uw', 'RS', 'IsInteger', 3, maybePCRead, maybePCWrite), + 'Rn': ('IntReg', 'uw', 'RN', 'IsInteger', 3, maybePCRead, maybePCWrite), + 'R7': ('IntReg', 'uw', '7', 'IsInteger', 3), + 'R0': ('IntReg', 'uw', '0', 'IsInteger', 3), - 'LR': ('IntReg', 'uw', 'INTREG_LR', 'IsInteger', 2), - 'CondCodes': ('IntReg', 'uw', 'INTREG_CONDCODES', None, 2), + 'LR': ('IntReg', 'uw', 'INTREG_LR', 'IsInteger', 3), + 'CondCodes': ('IntReg', 'uw', 'INTREG_CONDCODES', None, 3), 'OptCondCodes': ('IntReg', 'uw', '''(condCode == COND_AL || condCode == COND_UC) ? - INTREG_ZERO : INTREG_CONDCODES''', None, 2), - 'FpCondCodes': ('IntReg', 'uw', 'INTREG_FPCONDCODES', None, 2), + INTREG_ZERO : INTREG_CONDCODES''', None, 3), + 'FpCondCodes': ('IntReg', 'uw', 'INTREG_FPCONDCODES', None, 3), #Register fields for microops - 'Ra' : ('IntReg', 'uw', 'ura', 'IsInteger', 2, maybePCRead, maybePCWrite), - 'IWRa' : ('IntReg', 'uw', 'ura', 'IsInteger', 2, + 'Ra' : ('IntReg', 'uw', 'ura', 'IsInteger', 3, maybePCRead, maybePCWrite), + 'IWRa' : ('IntReg', 'uw', 'ura', 'IsInteger', 3, maybePCRead, maybeIWPCWrite), - 'Fa' : ('FloatReg', 'sf', 'ura', 'IsFloating', 2), - 'Rb' : ('IntReg', 'uw', 'urb', 'IsInteger', 2, maybePCRead, maybePCWrite), - 'Rc' : ('IntReg', 'uw', 'urc', 'IsInteger', 2, maybePCRead, maybePCWrite), + 'Fa' : ('FloatReg', 'sf', 'ura', 'IsFloating', 3), + 'Rb' : ('IntReg', 'uw', 'urb', 'IsInteger', 3, maybePCRead, maybePCWrite), + 'Rc' : ('IntReg', 'uw', 'urc', 'IsInteger', 3, maybePCRead, maybePCWrite), #General Purpose Floating Point Reg Operands - 'Fd': ('FloatReg', 'df', 'FD', 'IsFloating', 2), - 'Fn': ('FloatReg', 'df', 'FN', 'IsFloating', 2), - 'Fm': ('FloatReg', 'df', 'FM', 'IsFloating', 2), + 'Fd': ('FloatReg', 'df', 'FD', 'IsFloating', 3), + 'Fn': ('FloatReg', 'df', 'FN', 'IsFloating', 3), + 'Fm': ('FloatReg', 'df', 'FM', 'IsFloating', 3), #Memory Operand - 'Mem': ('Mem', 'uw', None, ('IsMemRef', 'IsLoad', 'IsStore'), 2), + 'Mem': ('Mem', 'uw', None, ('IsMemRef', 'IsLoad', 'IsStore'), 3), - 'Cpsr': ('ControlReg', 'uw', 'MISCREG_CPSR', (None, None, 'IsControl'), 1), - 'Itstate': ('ControlReg', 'ub', 'MISCREG_ITSTATE', None, 2), - 'Spsr': ('ControlReg', 'uw', 'MISCREG_SPSR', None, 2), - 'Fpsr': ('ControlReg', 'uw', 'MISCREG_FPSR', None, 2), - 'Fpsid': ('ControlReg', 'uw', 'MISCREG_FPSID', None, 2), - 'Fpscr': ('ControlReg', 'uw', 'MISCREG_FPSCR', None, 2), - 'Cpacr': ('ControlReg', 'uw', 'MISCREG_CPACR', (None, None, 'IsControl'), 2), - 'Fpexc': ('ControlReg', 'uw', 'MISCREG_FPEXC', None, 2), - 'Sctlr': ('ControlReg', 'uw', 'MISCREG_SCTLR', None, 2), - 'SevMailbox': ('ControlReg', 'uw', 'MISCREG_SEV_MAILBOX', None, 2), - 'PC': ('PC', 'ud', None, None, 2), - 'NPC': ('NPC', 'ud', None, (None, None, 'IsControl'), 2, - readNPC, writeNPC), - 'FNPC': ('NPC', 'ud', None, (None, None, 'IsControl'), 2, - readNPC, forceNPC), - 'IWNPC': ('NPC', 'ud', None, (None, None, 'IsControl'), 2, - readNPC, writeIWNPC), + 'Cpsr': ('ControlReg', 'uw', 'MISCREG_CPSR', (None, None, 'IsControl'), 2), + 'Itstate': ('ControlReg', 'ub', 'MISCREG_ITSTATE', None, 3), + 'Spsr': ('ControlReg', 'uw', 'MISCREG_SPSR', None, 3), + 'Fpsr': ('ControlReg', 'uw', 'MISCREG_FPSR', None, 3), + 'Fpsid': ('ControlReg', 'uw', 'MISCREG_FPSID', None, 3), + 'Fpscr': ('ControlReg', 'uw', 'MISCREG_FPSCR', None, 3), + 'Cpacr': ('ControlReg', 'uw', 'MISCREG_CPACR', (None, None, 'IsControl'), 3), + 'Fpexc': ('ControlReg', 'uw', 'MISCREG_FPEXC', None, 3), + 'Sctlr': ('ControlReg', 'uw', 'MISCREG_SCTLR', None, 3), + 'SevMailbox': ('ControlReg', 'uw', 'MISCREG_SEV_MAILBOX', None, 3), + #PCS needs to have a sorting index (the number at the end) less than all + #the integer registers which might update the PC. That way if the flag + #bits of the pc state are updated and a branch happens through R15, the + #updates are layered properly and the R15 update isn't lost. + 'PCS': ('PCState', 'uw', None, (None, None, 'IsControl'), 0) }}; diff --git a/src/arch/arm/isa_traits.hh b/src/arch/arm/isa_traits.hh index 8d3f0ffe3..f6aa7fcf0 100644 --- a/src/arch/arm/isa_traits.hh +++ b/src/arch/arm/isa_traits.hh @@ -123,15 +123,6 @@ namespace ArmISA INT_FIQ, NumInterruptTypes }; - - // These otherwise unused bits of the PC are used to select a mode - // like the J and T bits of the CPSR. - static const Addr PcJBitShift = 33; - static const Addr PcJBit = ULL(1) << PcJBitShift; - static const Addr PcTBitShift = 34; - static const Addr PcTBit = ULL(1) << PcTBitShift; - static const Addr PcModeMask = (ULL(1) << PcJBitShift) | - (ULL(1) << PcTBitShift); }; using namespace ArmISA; diff --git a/src/arch/arm/linux/system.cc b/src/arch/arm/linux/system.cc index 7c4e00921..40658976b 100644 --- a/src/arch/arm/linux/system.cc +++ b/src/arch/arm/linux/system.cc @@ -103,8 +103,7 @@ LinuxArmSystem::startup() ThreadContext *tc = threadContexts[0]; // Set the initial PC to be at start of the kernel code - tc->setPC(tc->getSystemPtr()->kernelEntry & loadAddrMask); - tc->setNextPC(tc->readPC() + sizeof(MachInst)); + tc->pcState(tc->getSystemPtr()->kernelEntry & loadAddrMask); // Setup the machine type tc->setIntReg(0, 0); diff --git a/src/arch/arm/nativetrace.cc b/src/arch/arm/nativetrace.cc index d97be88a2..75546f8de 100644 --- a/src/arch/arm/nativetrace.cc +++ b/src/arch/arm/nativetrace.cc @@ -106,7 +106,7 @@ Trace::ArmNativeTrace::ThreadState::update(ThreadContext *tc) } //R15, aliased with the PC - newState[STATE_PC] = tc->readNextPC(); + newState[STATE_PC] = tc->pcState().npc(); changed[STATE_PC] = (newState[STATE_PC] != oldState[STATE_PC]); //CPSR @@ -121,7 +121,7 @@ Trace::ArmNativeTrace::check(NativeTraceRecord *record) ThreadContext *tc = record->getThread(); // This area is read only on the target. It can't stop there to tell us // what's going on, so we should skip over anything there also. - if (tc->readNextPC() > 0xffff0000) + if (tc->nextInstAddr() > 0xffff0000) return; nState.update(this); mState.update(tc); diff --git a/src/arch/arm/predecoder.cc b/src/arch/arm/predecoder.cc index 04cec59b9..456b9e4c4 100644 --- a/src/arch/arm/predecoder.cc +++ b/src/arch/arm/predecoder.cc @@ -148,11 +148,11 @@ Predecoder::process() //Use this to give data to the predecoder. This should be used //when there is control flow. void -Predecoder::moreBytes(Addr pc, Addr fetchPC, MachInst inst) +Predecoder::moreBytes(const PCState &pc, Addr fetchPC, MachInst inst) { data = inst; - offset = (fetchPC >= pc) ? 0 : pc - fetchPC; - emi.thumb = isThumb(pc); + offset = (fetchPC >= pc.instAddr()) ? 0 : pc.instAddr() - fetchPC; + emi.thumb = pc.thumb(); FPSCR fpscr = tc->readMiscReg(MISCREG_FPSCR); emi.fpscrLen = fpscr.len; emi.fpscrStride = fpscr.stride; diff --git a/src/arch/arm/predecoder.hh b/src/arch/arm/predecoder.hh index 2db550024..47242b8ff 100644 --- a/src/arch/arm/predecoder.hh +++ b/src/arch/arm/predecoder.hh @@ -94,7 +94,7 @@ namespace ArmISA //Use this to give data to the predecoder. This should be used //when there is control flow. - void moreBytes(Addr pc, Addr fetchPC, MachInst inst); + void moreBytes(const PCState &pc, Addr fetchPC, MachInst inst); //Use this to give data to the predecoder. This should be used //when instructions are executed in order. @@ -121,9 +121,10 @@ namespace ArmISA } //This returns a constant reference to the ExtMachInst to avoid a copy - ExtMachInst getExtMachInst() + ExtMachInst getExtMachInst(PCState &pc) { ExtMachInst thisEmi = emi; + pc.npc(pc.pc() + getInstSize()); emi = 0; return thisEmi; } diff --git a/src/arch/arm/process.cc b/src/arch/arm/process.cc index bc2aee4c8..a5460ac19 100644 --- a/src/arch/arm/process.cc +++ b/src/arch/arm/process.cc @@ -360,11 +360,11 @@ ArmLiveProcess::argsInit(int intSize, int pageSize) tc->setIntReg(ArgumentReg2, 0); } - Addr prog_entry = objFile->entryPoint(); - if (arch == ObjectFile::Thumb) - prog_entry = (prog_entry & ~mask(1)) | PcTBit; - tc->setPC(prog_entry); - tc->setNextPC(prog_entry + sizeof(MachInst)); + PCState pc; + pc.thumb(arch == ObjectFile::Thumb); + pc.nextThumb(pc.thumb()); + pc.set(objFile->entryPoint() & ~mask(1)); + tc->pcState(pc); //Align the "stack_min" to a page boundary. stack_min = roundDown(stack_min, pageSize); diff --git a/src/arch/arm/system.hh b/src/arch/arm/system.hh index fe5ba6447..c64673df5 100644 --- a/src/arch/arm/system.hh +++ b/src/arch/arm/system.hh @@ -70,7 +70,7 @@ class ArmSystem : public System // Remove the low bit that thumb symbols have set // but that aren't actually odd aligned if (addr & 0x1) - return (addr & ~1) | PcTBit; + return addr & ~1; return addr; } }; diff --git a/src/arch/arm/table_walker.cc b/src/arch/arm/table_walker.cc index 73dd24e1c..06bb10219 100644 --- a/src/arch/arm/table_walker.cc +++ b/src/arch/arm/table_walker.cc @@ -107,7 +107,7 @@ TableWalker::walk(RequestPtr _req, ThreadContext *_tc, uint8_t _cid, TLB::Mode _ /** @todo These should be cached or grabbed from cached copies in the TLB, all these miscreg reads are expensive */ - currState->vaddr = currState->req->getVaddr() & ~PcModeMask; + currState->vaddr = currState->req->getVaddr(); currState->sctlr = currState->tc->readMiscReg(MISCREG_SCTLR); sctlr = currState->sctlr; currState->N = currState->tc->readMiscReg(MISCREG_TTBCR); diff --git a/src/arch/arm/tlb.cc b/src/arch/arm/tlb.cc index c0ebb52b2..239d5d8a2 100644 --- a/src/arch/arm/tlb.cc +++ b/src/arch/arm/tlb.cc @@ -316,7 +316,7 @@ TLB::translateSe(RequestPtr req, ThreadContext *tc, Mode mode, Translation *translation, bool &delay, bool timing) { // XXX Cache misc registers and have miscreg write function inv cache - Addr vaddr = req->getVaddr() & ~PcModeMask; + Addr vaddr = req->getVaddr(); SCTLR sctlr = tc->readMiscReg(MISCREG_SCTLR); uint32_t flags = req->getFlags(); @@ -362,7 +362,7 @@ TLB::translateFs(RequestPtr req, ThreadContext *tc, Mode mode, Translation *translation, bool &delay, bool timing) { // XXX Cache misc registers and have miscreg write function inv cache - Addr vaddr = req->getVaddr() & ~PcModeMask; + Addr vaddr = req->getVaddr(); SCTLR sctlr = tc->readMiscReg(MISCREG_SCTLR); CPSR cpsr = tc->readMiscReg(MISCREG_CPSR); uint32_t flags = req->getFlags(); diff --git a/src/arch/arm/types.hh b/src/arch/arm/types.hh index 3c3b29494..57f34e3c2 100644 --- a/src/arch/arm/types.hh +++ b/src/arch/arm/types.hh @@ -43,8 +43,10 @@ #ifndef __ARCH_ARM_TYPES_HH__ #define __ARCH_ARM_TYPES_HH__ +#include "arch/generic/types.hh" #include "base/bitunion.hh" #include "base/hashmap.hh" +#include "base/misc.hh" #include "base/types.hh" namespace ArmISA @@ -188,6 +190,189 @@ namespace ArmISA Bitfield<11, 8> ltcoproc; EndBitUnion(ExtMachInst) + class PCState : public GenericISA::UPCState + { + protected: + + typedef GenericISA::UPCState Base; + + enum FlagBits { + ThumbBit = (1 << 0), + JazelleBit = (1 << 1) + }; + uint8_t flags; + uint8_t nextFlags; + + public: + PCState() : flags(0), nextFlags(0) + {} + + void + set(Addr val) + { + Base::set(val); + npc(val + (thumb() ? 2 : 4)); + } + + PCState(Addr val) : flags(0), nextFlags(0) + { set(val); } + + bool + thumb() const + { + return flags & ThumbBit; + } + + void + thumb(bool val) + { + if (val) + flags |= ThumbBit; + else + flags &= ~ThumbBit; + } + + bool + nextThumb() const + { + return nextFlags & ThumbBit; + } + + void + nextThumb(bool val) + { + if (val) + nextFlags |= ThumbBit; + else + nextFlags &= ~ThumbBit; + } + + bool + jazelle() const + { + return flags & JazelleBit; + } + + void + jazelle(bool val) + { + if (val) + flags |= JazelleBit; + else + flags &= ~JazelleBit; + } + + bool + nextJazelle() const + { + return nextFlags & JazelleBit; + } + + void + nextJazelle(bool val) + { + if (val) + nextFlags |= JazelleBit; + else + nextFlags &= ~JazelleBit; + } + + void + advance() + { + Base::advance(); + npc(pc() + (thumb() ? 2 : 4)); + flags = nextFlags; + } + + void + uEnd() + { + advance(); + upc(0); + nupc(1); + } + + Addr + instPC() const + { + return pc() + (thumb() ? 4 : 8); + } + + void + instNPC(uint32_t val) + { + npc(val &~ mask(nextThumb() ? 1 : 2)); + } + + Addr + instNPC() const + { + return npc(); + } + + // Perform an interworking branch. + void + instIWNPC(uint32_t val) + { + bool thumbEE = (thumb() && jazelle()); + + Addr newPC = val; + if (thumbEE) { + if (bits(newPC, 0)) { + newPC = newPC & ~mask(1); + } else { + panic("Bad thumbEE interworking branch address %#x.\n", + newPC); + } + } else { + if (bits(newPC, 0)) { + nextThumb(true); + newPC = newPC & ~mask(1); + } else if (!bits(newPC, 1)) { + nextThumb(false); + } else { + warn("Bad interworking branch address %#x.\n", newPC); + } + } + npc(newPC); + } + + // Perform an interworking branch in ARM mode, a regular branch + // otherwise. + void + instAIWNPC(uint32_t val) + { + if (!thumb() && !jazelle()) + instIWNPC(val); + else + instNPC(val); + } + + bool + operator == (const PCState &opc) const + { + return Base::operator == (opc) && + flags == opc.flags && nextFlags == opc.nextFlags; + } + + void + serialize(std::ostream &os) + { + Base::serialize(os); + SERIALIZE_SCALAR(flags); + SERIALIZE_SCALAR(nextFlags); + } + + void + unserialize(Checkpoint *cp, const std::string §ion) + { + Base::unserialize(cp, section); + UNSERIALIZE_SCALAR(flags); + UNSERIALIZE_SCALAR(nextFlags); + } + }; + // Shift types for ARM instructions enum ArmShiftType { LSL = 0, diff --git a/src/arch/arm/utility.cc b/src/arch/arm/utility.cc index 7609b3991..a114ec5e0 100644 --- a/src/arch/arm/utility.cc +++ b/src/arch/arm/utility.cc @@ -128,13 +128,9 @@ readCp15Register(uint32_t &Rd, int CRn, int opc1, int CRm, int opc2) void skipFunction(ThreadContext *tc) { - Addr newpc = tc->readIntReg(ReturnAddressReg); - newpc &= ~ULL(1); - if (isThumb(tc->readPC())) - tc->setPC(newpc | PcTBit); - else - tc->setPC(newpc); - tc->setNextPC(tc->readPC() + sizeof(TheISA::MachInst)); + TheISA::PCState newPC = tc->pcState(); + newPC.set(tc->readIntReg(ReturnAddressReg) & ~ULL(1)); + tc->pcState(newPC); } diff --git a/src/arch/arm/utility.hh b/src/arch/arm/utility.hh index 571a74ef8..a92ab072c 100644 --- a/src/arch/arm/utility.hh +++ b/src/arch/arm/utility.hh @@ -51,10 +51,19 @@ #include "base/misc.hh" #include "base/trace.hh" #include "base/types.hh" +#include "cpu/static_inst.hh" #include "cpu/thread_context.hh" namespace ArmISA { + inline PCState + buildRetPC(const PCState &curPC, const PCState &callPC) + { + PCState retPC = callPC; + retPC.uEnd(); + return retPC; + } + inline bool testPredicate(CPSR cpsr, ConditionCode code) { @@ -93,12 +102,6 @@ namespace ArmISA { tc->activate(0); } - static inline bool - isThumb(Addr pc) - { - return (pc & PcTBit); - } - static inline void copyRegs(ThreadContext *src, ThreadContext *dest) { @@ -163,6 +166,12 @@ Fault readCp15Register(uint32_t &Rd, int CRn, int opc1, int CRm, int opc2); void skipFunction(ThreadContext *tc); +inline void +advancePC(PCState &pc, const StaticInstPtr inst) +{ + inst->advancePC(pc); +} + }; diff --git a/src/arch/generic/types.hh b/src/arch/generic/types.hh new file mode 100644 index 000000000..214b01926 --- /dev/null +++ b/src/arch/generic/types.hh @@ -0,0 +1,432 @@ +/* + * Copyright (c) 2010 Gabe Black + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are + * met: redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer; + * redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution; + * neither the name of the copyright holders nor the names of its + * contributors may be used to endorse or promote products derived from + * this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT + * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, + * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT + * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE + * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + * + * Authors: Gabe Black + */ + +#ifndef __ARCH_GENERIC_TYPES_HH__ +#define __ARCH_GENERIC_TYPES_HH__ + +#include + +#include "base/types.hh" +#include "base/trace.hh" +#include "sim/serialize.hh" + +namespace GenericISA +{ + +// The guaranteed interface. +class PCStateBase +{ + protected: + Addr _pc; + Addr _npc; + + PCStateBase() {} + PCStateBase(Addr val) { set(val); } + + public: + /** + * Returns the memory address the bytes of this instruction came from. + * + * @return Memory address of the current instruction's encoding. + */ + Addr + instAddr() const + { + return _pc; + } + + /** + * Returns the memory address the bytes of the next instruction came from. + * + * @return Memory address of the next instruction's encoding. + */ + Addr + nextInstAddr() const + { + return _npc; + } + + /** + * Returns the current micropc. + * + * @return The current micropc. + */ + MicroPC + microPC() const + { + return 0; + } + + /** + * Force this PC to reflect a particular value, resetting all its other + * fields around it. This is useful for in place (re)initialization. + * + * @param val The value to set the PC to. + */ + void set(Addr val); + + bool + operator == (const PCStateBase &opc) const + { + return _pc == opc._pc && _npc == opc._npc; + } + + void + serialize(std::ostream &os) + { + SERIALIZE_SCALAR(_pc); + SERIALIZE_SCALAR(_npc); + } + + void + unserialize(Checkpoint *cp, const std::string §ion) + { + UNSERIALIZE_SCALAR(_pc); + UNSERIALIZE_SCALAR(_npc); + } +}; + + +/* + * Different flavors of PC state. Only ISA specific code should rely on + * any particular type of PC state being available. All other code should + * use the interface above. + */ + +// The most basic type of PC. +template +class SimplePCState : public PCStateBase +{ + protected: + typedef PCStateBase Base; + + public: + + Addr pc() const { return _pc; } + void pc(Addr val) { _pc = val; } + + Addr npc() const { return _npc; } + void npc(Addr val) { _npc = val; } + + void + set(Addr val) + { + pc(val); + npc(val + sizeof(MachInst)); + }; + + SimplePCState() {} + SimplePCState(Addr val) { set(val); } + + bool + branching() const + { + return this->npc() != this->pc() + sizeof(MachInst); + } + + // Advance the PC. + void + advance() + { + _pc = _npc; + _npc += sizeof(MachInst); + } +}; + +template +std::ostream & +operator<<(std::ostream & os, const SimplePCState &pc) +{ + ccprintf(os, "(%#x=>%#x)", pc.pc(), pc.npc()); + return os; +} + +// A PC and microcode PC. +template +class UPCState : public SimplePCState +{ + protected: + typedef SimplePCState Base; + + MicroPC _upc; + MicroPC _nupc; + + public: + + MicroPC upc() const { return _upc; } + void upc(MicroPC val) { _upc = val; } + + MicroPC nupc() const { return _nupc; } + void nupc(MicroPC val) { _nupc = val; } + + MicroPC + microPC() const + { + return _upc; + } + + void + set(Addr val) + { + Base::set(val); + upc(0); + nupc(1); + } + + UPCState() {} + UPCState(Addr val) { set(val); } + + bool + branching() const + { + return this->npc() != this->pc() + sizeof(MachInst) || + this->nupc() != this->upc() + 1; + } + + // Advance the upc within the instruction. + void + uAdvance() + { + _upc = _nupc; + _nupc++; + } + + // End the macroop by resetting the upc and advancing the regular pc. + void + uEnd() + { + this->advance(); + _upc = 0; + _nupc = 1; + } + + bool + operator == (const UPCState &opc) const + { + return Base::_pc == opc._pc && + Base::_npc == opc._npc && + _upc == opc._upc && _nupc == opc._nupc; + } + + void + serialize(std::ostream &os) + { + Base::serialize(os); + SERIALIZE_SCALAR(_upc); + SERIALIZE_SCALAR(_nupc); + } + + void + unserialize(Checkpoint *cp, const std::string §ion) + { + Base::unserialize(cp, section); + UNSERIALIZE_SCALAR(_upc); + UNSERIALIZE_SCALAR(_nupc); + } +}; + +template +std::ostream & +operator<<(std::ostream & os, const UPCState &pc) +{ + ccprintf(os, "(%#x=>%#x).(%d=>%d)", + pc.pc(), pc.npc(), pc.upc(), pc.npc()); + return os; +} + +// A PC with a delay slot. +template +class DelaySlotPCState : public SimplePCState +{ + protected: + typedef SimplePCState Base; + + Addr _nnpc; + + public: + + Addr nnpc() const { return _nnpc; } + void nnpc(Addr val) { _nnpc = val; } + + void + set(Addr val) + { + Base::set(val); + nnpc(val + 2 * sizeof(MachInst)); + } + + DelaySlotPCState() {} + DelaySlotPCState(Addr val) { set(val); } + + bool + branching() const + { + return !(this->nnpc() == this->npc() + sizeof(MachInst) && + (this->npc() == this->pc() + sizeof(MachInst) || + this->npc() == this->pc() + 2 * sizeof(MachInst))); + } + + // Advance the PC. + void + advance() + { + Base::_pc = Base::_npc; + Base::_npc = _nnpc; + _nnpc += sizeof(MachInst); + } + + bool + operator == (const DelaySlotPCState &opc) const + { + return Base::_pc == opc._pc && + Base::_npc == opc._npc && + _nnpc == opc._nnpc; + } + + void + serialize(std::ostream &os) + { + Base::serialize(os); + SERIALIZE_SCALAR(_nnpc); + } + + void + unserialize(Checkpoint *cp, const std::string §ion) + { + Base::unserialize(cp, section); + UNSERIALIZE_SCALAR(_nnpc); + } +}; + +template +std::ostream & +operator<<(std::ostream & os, const DelaySlotPCState &pc) +{ + ccprintf(os, "(%#x=>%#x=>%#x)", + pc.pc(), pc.npc(), pc.nnpc()); + return os; +} + +// A PC with a delay slot and a microcode PC. +template +class DelaySlotUPCState : public DelaySlotPCState +{ + protected: + typedef DelaySlotPCState Base; + + MicroPC _upc; + MicroPC _nupc; + + public: + + MicroPC upc() const { return _upc; } + void upc(MicroPC val) { _upc = val; } + + MicroPC nupc() const { return _nupc; } + void nupc(MicroPC val) { _nupc = val; } + + MicroPC + microPC() const + { + return _upc; + } + + void + set(Addr val) + { + Base::set(val); + upc(0); + nupc(1); + } + + DelaySlotUPCState() {} + DelaySlotUPCState(Addr val) { set(val); } + + bool + branching() const + { + return Base::branching() || this->nupc() != this->upc() + 1; + } + + // Advance the upc within the instruction. + void + uAdvance() + { + _upc = _nupc; + _nupc++; + } + + // End the macroop by resetting the upc and advancing the regular pc. + void + uEnd() + { + this->advance(); + _upc = 0; + _nupc = 1; + } + + bool + operator == (const DelaySlotUPCState &opc) const + { + return Base::_pc == opc._pc && + Base::_npc == opc._npc && + Base::_nnpc == opc._nnpc && + _upc == opc._upc && _nupc == opc._nupc; + } + + void + serialize(std::ostream &os) + { + Base::serialize(os); + SERIALIZE_SCALAR(_upc); + SERIALIZE_SCALAR(_nupc); + } + + void + unserialize(Checkpoint *cp, const std::string §ion) + { + Base::unserialize(cp, section); + UNSERIALIZE_SCALAR(_upc); + UNSERIALIZE_SCALAR(_nupc); + } +}; + +template +std::ostream & +operator<<(std::ostream & os, const DelaySlotUPCState &pc) +{ + ccprintf(os, "(%#x=>%#x=>%#x).(%d=>%d)", + pc.pc(), pc.npc(), pc.nnpc(), pc.upc(), pc.nupc()); + return os; +} + +} + +#endif diff --git a/src/arch/isa_parser.py b/src/arch/isa_parser.py index 4e06c2ded..8e13b6a6a 100755 --- a/src/arch/isa_parser.py +++ b/src/arch/isa_parser.py @@ -681,71 +681,43 @@ class MemOperand(Operand): def makeAccSize(self): return self.size -class PCOperand(Operand): +class PCStateOperand(Operand): def makeConstructor(self): return '' def makeRead(self): - return '%s = xc->readPC();\n' % self.base_name + return '%s = xc->pcState();\n' % self.base_name def makeWrite(self): - return 'xc->setPC(%s);\n' % self.base_name + return 'xc->pcState(%s);\n' % self.base_name -class UPCOperand(Operand): - def makeConstructor(self): - return '' - - def makeRead(self): - if self.read_code != None: - return self.buildReadCode('readMicroPC') - return '%s = xc->readMicroPC();\n' % self.base_name - - def makeWrite(self): - if self.write_code != None: - return self.buildWriteCode('setMicroPC') - return 'xc->setMicroPC(%s);\n' % self.base_name + def makeDecl(self): + return 'TheISA::PCState ' + self.base_name + ' M5_VAR_USED;\n'; -class NUPCOperand(Operand): +class PCOperand(Operand): def makeConstructor(self): return '' def makeRead(self): - if self.read_code != None: - return self.buildReadCode('readNextMicroPC') - return '%s = xc->readNextMicroPC();\n' % self.base_name + return '%s = xc->instAddr();\n' % self.base_name - def makeWrite(self): - if self.write_code != None: - return self.buildWriteCode('setNextMicroPC') - return 'xc->setNextMicroPC(%s);\n' % self.base_name - -class NPCOperand(Operand): +class UPCOperand(Operand): def makeConstructor(self): return '' def makeRead(self): if self.read_code != None: - return self.buildReadCode('readNextPC') - return '%s = xc->readNextPC();\n' % self.base_name - - def makeWrite(self): - if self.write_code != None: - return self.buildWriteCode('setNextPC') - return 'xc->setNextPC(%s);\n' % self.base_name + return self.buildReadCode('microPC') + return '%s = xc->microPC();\n' % self.base_name -class NNPCOperand(Operand): +class NPCOperand(Operand): def makeConstructor(self): return '' def makeRead(self): if self.read_code != None: - return self.buildReadCode('readNextNPC') - return '%s = xc->readNextNPC();\n' % self.base_name - - def makeWrite(self): - if self.write_code != None: - return self.buildWriteCode('setNextNPC') - return 'xc->setNextNPC(%s);\n' % self.base_name + return self.buildReadCode('nextInstAddr') + return '%s = xc->nextInstAddr();\n' % self.base_name class OperandList(object): '''Find all the operands in the given code block. Returns an operand diff --git a/src/arch/mips/isa/base.isa b/src/arch/mips/isa/base.isa index 4e2b12fc4..cd6faf0f3 100644 --- a/src/arch/mips/isa/base.isa +++ b/src/arch/mips/isa/base.isa @@ -56,6 +56,13 @@ output header {{ void printReg(std::ostream &os, int reg) const; std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const; + + public: + void + advancePC(MipsISA::PCState &pc) const + { + pc.advance(); + } }; }}; diff --git a/src/arch/mips/isa/decoder.isa b/src/arch/mips/isa/decoder.isa index 9832937b5..3f3bd370a 100644 --- a/src/arch/mips/isa/decoder.isa +++ b/src/arch/mips/isa/decoder.isa @@ -133,25 +133,34 @@ decode OPCODE_HI default Unknown::unknown() { 0x1: jr_hb({{ Config1Reg config1 = Config1; if (config1.ca == 0) { - NNPC = Rs; + pc.nnpc(Rs); } else { panic("MIPS16e not supported\n"); } + PCS = pc; }}, IsReturn, ClearHazards); default: jr({{ Config1Reg config1 = Config1; if (config1.ca == 0) { - NNPC = Rs; + pc.nnpc(Rs); } else { panic("MIPS16e not supported\n"); } + PCS = pc; }}, IsReturn); } 0x1: decode HINT { - 0x1: jalr_hb({{ Rd = NNPC; NNPC = Rs; }}, IsCall - , ClearHazards); - default: jalr({{ Rd = NNPC; NNPC = Rs; }}, IsCall); + 0x1: jalr_hb({{ + Rd = pc.nnpc(); + pc.nnpc(Rs); + PCS = pc; + }}, IsCall, ClearHazards); + default: jalr({{ + Rd = pc.nnpc(); + pc.nnpc(Rs); + PCS = pc; + }}, IsCall); } } @@ -323,9 +332,14 @@ decode OPCODE_HI default Unknown::unknown() { } format Jump { - 0x2: j({{ NNPC = (NPC & 0xF0000000) | (JMPTARG << 2); }}); - 0x3: jal({{ NNPC = (NPC & 0xF0000000) | (JMPTARG << 2); }}, - IsCall, Link); + 0x2: j({{ + pc.nnpc((pc.npc() & 0xF0000000) | (JMPTARG << 2)); + PCS = pc; + }}); + 0x3: jal({{ + pc.nnpc((pc.npc() & 0xF0000000) | (JMPTARG << 2)); + PCS = pc; + }}, IsCall, Link); } format Branch { @@ -694,15 +708,16 @@ decode OPCODE_HI default Unknown::unknown() { ConfigReg config = Config; SRSCtlReg srsCtl = SRSCtl; DPRINTF(MipsPRA,"Restoring PC - %x\n",EPC); + MipsISA::PCState pc = PCS; if (status.erl == 1) { status.erl = 0; - NPC = ErrorEPC; + pc.npc(ErrorEPC); // Need to adjust NNPC, otherwise things break - NNPC = ErrorEPC + sizeof(MachInst); + pc.nnpc(ErrorEPC + sizeof(MachInst)); } else { - NPC = EPC; + pc.npc(EPC); // Need to adjust NNPC, otherwise things break - NNPC = EPC + sizeof(MachInst); + pc.nnpc(EPC + sizeof(MachInst)); status.exl = 0; if (config.ar >=1 && srsCtl.hss > 0 && @@ -711,6 +726,7 @@ decode OPCODE_HI default Unknown::unknown() { //xc->setShadowSet(srsCtl.pss); } } + PCS = pc; LLFlag = 0; Status = status; SRSCtl = srsCtl; @@ -718,13 +734,15 @@ decode OPCODE_HI default Unknown::unknown() { 0x1F: deret({{ DebugReg debug = Debug; + MipsISA::PCState pc = PCS; if (debug.dm == 1) { debug.dm = 1; debug.iexi = 0; - NPC = DEPC; + pc.npc(DEPC); } else { // Undefined; } + PCS = pc; Debug = debug; }}, IsReturn, IsSerializing, IsERET); } diff --git a/src/arch/mips/isa/formats/branch.isa b/src/arch/mips/isa/formats/branch.isa index 78f973a70..232a743a7 100644 --- a/src/arch/mips/isa/formats/branch.isa +++ b/src/arch/mips/isa/formats/branch.isa @@ -89,7 +89,7 @@ output header {{ } } - Addr branchTarget(Addr branchPC) const; + MipsISA::PCState branchTarget(const MipsISA::PCState &branchPC) const; std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const; @@ -116,7 +116,7 @@ output header {{ { } - Addr branchTarget(ThreadContext *tc) const; + MipsISA::PCState branchTarget(ThreadContext *tc) const; std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const; @@ -124,17 +124,25 @@ output header {{ }}; output decoder {{ - Addr - Branch::branchTarget(Addr branchPC) const + MipsISA::PCState + Branch::branchTarget(const MipsISA::PCState &branchPC) const { - return branchPC + 4 + disp; + MipsISA::PCState target = branchPC; + target.advance(); + target.npc(branchPC.pc() + sizeof(MachInst) + disp); + target.nnpc(target.npc() + sizeof(MachInst)); + return target; } - Addr + MipsISA::PCState Jump::branchTarget(ThreadContext *tc) const { - Addr NPC = tc->readNextPC(); - return (NPC & 0xF0000000) | (disp); + MipsISA::PCState target = tc->pcState(); + Addr pc = target.pc(); + target.advance(); + target.npc((pc & 0xF0000000) | disp); + target.nnpc(target.npc() + sizeof(MachInst)); + return target; } const std::string & @@ -217,19 +225,16 @@ output decoder {{ }}; def format Branch(code, *opt_flags) {{ - not_taken_code = ' NNPC = NNPC;\n' - not_taken_code += '} \n' + not_taken_code = '' #Build Instruction Flags #Use Link & Likely Flags to Add Link/Condition Code inst_flags = ('IsDirectControl', ) for x in opt_flags: if x == 'Link': - code += 'R31 = NNPC;\n' + code += 'R31 = pc.nnpc();\n' elif x == 'Likely': - not_taken_code = ' NPC = NNPC;\n' - not_taken_code += ' NNPC = NNPC + 4;\n' - not_taken_code += '} \n' + not_taken_code = 'pc.advance();' inst_flags += ('IsCondDelaySlot', ) else: inst_flags += (x, ) @@ -241,11 +246,17 @@ def format Branch(code, *opt_flags) {{ inst_flags += ('IsCondControl', ) #Condition code - code = 'bool cond;\n' + code - code += 'if (cond) {\n' - code += ' NNPC = NPC + disp;\n' - code += '} else {\n' - code += not_taken_code + code = ''' + bool cond; + MipsISA::PCState pc = PCS; + %(code)s + if (cond) { + pc.nnpc(pc.npc() + disp); + } else { + %(not_taken_code)s + } + PCS = pc; + ''' % { "code" : code, "not_taken_code" : not_taken_code } iop = InstObjParams(name, Name, 'Branch', code, inst_flags) header_output = BasicDeclare.subst(iop) @@ -255,19 +266,16 @@ def format Branch(code, *opt_flags) {{ }}; def format DspBranch(code, *opt_flags) {{ - not_taken_code = ' NNPC = NNPC;\n' - not_taken_code += '} \n' + not_taken_code = '' #Build Instruction Flags #Use Link & Likely Flags to Add Link/Condition Code inst_flags = ('IsDirectControl', ) for x in opt_flags: if x == 'Link': - code += 'R31 = NNPC;\n' + code += 'R32 = pc.nnpc();' elif x == 'Likely': - not_taken_code = ' NPC = NNPC;\n' - not_taken_code += ' NNPC = NNPC + 4;\n' - not_taken_code += '} \n' + not_taken_code = 'pc.advance();' inst_flags += ('IsCondDelaySlot', ) else: inst_flags += (x, ) @@ -278,19 +286,19 @@ def format DspBranch(code, *opt_flags) {{ else: inst_flags += ('IsCondControl', ) - #Declaration code - decl_code = 'bool cond;\n' - decl_code += 'uint32_t dspctl;\n' - - #Fetch code - fetch_code = 'dspctl = DSPControl;\n' - #Condition code - code = decl_code + fetch_code + code - code += 'if (cond) {\n' - code += ' NNPC = NPC + disp;\n' - code += '} else {\n' - code += not_taken_code + code = ''' + MipsISA::PCState pc = PCS; + bool cond; + uint32_t dspctl = DSPControl; + %(code)s + if (cond) { + pc.nnpc(pc.npc() + disp); + } else { + %(not_taken_code)s + } + PCS = pc; + ''' % { "code" : code, "not_taken_code" : not_taken_code } iop = InstObjParams(name, Name, 'Branch', code, inst_flags) header_output = BasicDeclare.subst(iop) @@ -305,12 +313,18 @@ def format Jump(code, *opt_flags) {{ inst_flags = ('IsIndirectControl', 'IsUncondControl') for x in opt_flags: if x == 'Link': - code = 'R31 = NNPC;\n' + code + code = ''' + R31 = pc.nnpc(); + ''' + code elif x == 'ClearHazards': code += '/* Code Needed to Clear Execute & Inst Hazards */\n' else: inst_flags += (x, ) + code = ''' + MipsISA::PCState pc = PCS; + ''' + code + iop = InstObjParams(name, Name, 'Jump', code, inst_flags) header_output = BasicDeclare.subst(iop) decoder_output = BasicConstructor.subst(iop) diff --git a/src/arch/mips/isa/includes.isa b/src/arch/mips/isa/includes.isa index 22eb3bf13..d5e1448ac 100644 --- a/src/arch/mips/isa/includes.isa +++ b/src/arch/mips/isa/includes.isa @@ -39,6 +39,7 @@ output header {{ #include #include "arch/mips/isa_traits.hh" +#include "arch/mips/types.hh" #include "cpu/static_inst.hh" #include "mem/packet.hh" }}; diff --git a/src/arch/mips/isa/operands.isa b/src/arch/mips/isa/operands.isa index 27cb4357a..1bb5ae5b3 100644 --- a/src/arch/mips/isa/operands.isa +++ b/src/arch/mips/isa/operands.isa @@ -151,6 +151,5 @@ def operands {{ 'Mem': ('Mem', 'uw', None, ('IsMemRef', 'IsLoad', 'IsStore'), 4), #Program Counter Operands - 'NPC': ('NPC', 'uw', None, 'IsControl', 4), - 'NNPC':('NNPC', 'uw', None, 'IsControl', 4) + 'PCS': ('PCState', 'uw', None, (None, None, 'IsControl'), 4) }}; diff --git a/src/arch/mips/mt.hh b/src/arch/mips/mt.hh index 7217c335e..3ec6cbe70 100755 --- a/src/arch/mips/mt.hh +++ b/src/arch/mips/mt.hh @@ -77,11 +77,12 @@ haltThread(TC *tc) // Save last known PC in TCRestart // @TODO: Needs to check if this is a branch and if so, // take previous instruction - tc->setMiscReg(MISCREG_TC_RESTART, tc->readNextPC()); + PCState pc = tc->pcState(); + tc->setMiscReg(MISCREG_TC_RESTART, pc.npc()); warn("%i: Halting thread %i in %s @ PC %x, setting restart PC to %x", curTick, tc->threadId(), tc->getCpuPtr()->name(), - tc->readPC(), tc->readNextPC()); + pc.pc(), pc.npc()); } } @@ -91,17 +92,14 @@ restoreThread(TC *tc) { if (tc->status() != TC::Active) { // Restore PC from TCRestart - IntReg pc = tc->readMiscRegNoEffect(MISCREG_TC_RESTART); + Addr restartPC = tc->readMiscRegNoEffect(MISCREG_TC_RESTART); // TODO: SET PC WITH AN EVENT INSTEAD OF INSTANTANEOUSLY - tc->setPC(pc); - tc->setNextPC(pc + 4); - tc->setNextNPC(pc + 8); + tc->pcState(restartPC); tc->activate(0); warn("%i: Restoring thread %i in %s @ PC %x", - curTick, tc->threadId(), tc->getCpuPtr()->name(), - tc->readPC()); + curTick, tc->threadId(), tc->getCpuPtr()->name(), restartPC); } } diff --git a/src/arch/mips/predecoder.hh b/src/arch/mips/predecoder.hh index c20fe1f5f..f059710e5 100644 --- a/src/arch/mips/predecoder.hh +++ b/src/arch/mips/predecoder.hh @@ -75,7 +75,7 @@ class Predecoder //Use this to give data to the predecoder. This should be used //when there is control flow. void - moreBytes(Addr pc, Addr fetchPC, MachInst inst) + moreBytes(const PCState &pc, Addr fetchPC, MachInst inst) { emi = inst; } @@ -94,7 +94,7 @@ class Predecoder //This returns a constant reference to the ExtMachInst to avoid a copy const ExtMachInst & - getExtMachInst() + getExtMachInst(PCState &pc) { return emi; } diff --git a/src/arch/mips/process.cc b/src/arch/mips/process.cc index fa3e3bff9..26a2a0ddb 100644 --- a/src/arch/mips/process.cc +++ b/src/arch/mips/process.cc @@ -176,10 +176,7 @@ MipsLiveProcess::argsInit(int pageSize) setSyscallArg(tc, 1, argv_array_base); tc->setIntReg(StackPointerReg, stack_min); - Addr prog_entry = objFile->entryPoint(); - tc->setPC(prog_entry); - tc->setNextPC(prog_entry + sizeof(MachInst)); - tc->setNextNPC(prog_entry + (2 * sizeof(MachInst))); + tc->pcState(objFile->entryPoint()); } diff --git a/src/arch/mips/types.hh b/src/arch/mips/types.hh index c7ef6afe1..f21db51b1 100644 --- a/src/arch/mips/types.hh +++ b/src/arch/mips/types.hh @@ -31,6 +31,7 @@ #ifndef __ARCH_MIPS_TYPES_HH__ #define __ARCH_MIPS_TYPES_HH__ +#include "arch/generic/types.hh" #include "base/types.hh" namespace MipsISA @@ -39,6 +40,8 @@ namespace MipsISA typedef uint32_t MachInst; typedef uint64_t ExtMachInst; +typedef GenericISA::DelaySlotPCState PCState; + typedef uint64_t LargestRead; //used in FP convert & round function diff --git a/src/arch/mips/utility.cc b/src/arch/mips/utility.cc index faae1e937..0859eb80f 100644 --- a/src/arch/mips/utility.cc +++ b/src/arch/mips/utility.cc @@ -267,10 +267,9 @@ copyMiscRegs(ThreadContext *src, ThreadContext *dest) void skipFunction(ThreadContext *tc) { - Addr newpc = tc->readIntReg(ReturnAddressReg); - tc->setPC(newpc); - tc->setNextPC(tc->readPC() + sizeof(TheISA::MachInst)); - tc->setNextPC(tc->readNextPC() + sizeof(TheISA::MachInst)); + TheISA::PCState newPC = tc->pcState(); + newPC.set(tc->readIntReg(ReturnAddressReg)); + tc->pcState(newPC); } diff --git a/src/arch/mips/utility.hh b/src/arch/mips/utility.hh index bc50027c0..2f6726c59 100644 --- a/src/arch/mips/utility.hh +++ b/src/arch/mips/utility.hh @@ -39,12 +39,22 @@ #include "base/misc.hh" #include "base/types.hh" #include "config/full_system.hh" +#include "cpu/static_inst.hh" #include "cpu/thread_context.hh" class ThreadContext; namespace MipsISA { +inline PCState +buildRetPC(const PCState &curPC, const PCState &callPC) +{ + PCState ret = callPC; + ret.advance(); + ret.pc(curPC.npc()); + return ret; +} + uint64_t getArgument(ThreadContext *tc, int &number, uint16_t size, bool fp); //////////////////////////////////////////////////////////////////////// @@ -105,6 +115,12 @@ void copyMiscRegs(ThreadContext *src, ThreadContext *dest); void skipFunction(ThreadContext *tc); +inline void +advancePC(PCState &pc, const StaticInstPtr inst) +{ + pc.advance(); +} + }; diff --git a/src/arch/power/insts/branch.cc b/src/arch/power/insts/branch.cc index c10f7c996..352c4ea57 100644 --- a/src/arch/power/insts/branch.cc +++ b/src/arch/power/insts/branch.cc @@ -52,10 +52,10 @@ PCDependentDisassembly::disassemble(Addr pc, const SymbolTable *symtab) const return *cachedDisassembly; } -Addr -BranchPCRel::branchTarget(Addr pc) const +PowerISA::PCState +BranchPCRel::branchTarget(const PowerISA::PCState &pc) const { - return (uint32_t)(pc + disp); + return (uint32_t)(pc.pc() + disp); } std::string @@ -76,8 +76,8 @@ BranchPCRel::generateDisassembly(Addr pc, const SymbolTable *symtab) const return ss.str(); } -Addr -BranchNonPCRel::branchTarget(Addr pc) const +PowerISA::PCState +BranchNonPCRel::branchTarget(const PowerISA::PCState &pc) const { return targetAddr; } @@ -98,10 +98,10 @@ BranchNonPCRel::generateDisassembly(Addr pc, const SymbolTable *symtab) const return ss.str(); } -Addr -BranchPCRelCond::branchTarget(Addr pc) const +PowerISA::PCState +BranchPCRelCond::branchTarget(const PowerISA::PCState &pc) const { - return (uint32_t)(pc + disp); + return (uint32_t)(pc.pc() + disp); } std::string @@ -124,8 +124,8 @@ BranchPCRelCond::generateDisassembly(Addr pc, const SymbolTable *symtab) const return ss.str(); } -Addr -BranchNonPCRelCond::branchTarget(Addr pc) const +PowerISA::PCState +BranchNonPCRelCond::branchTarget(const PowerISA::PCState &pc) const { return targetAddr; } @@ -149,11 +149,11 @@ BranchNonPCRelCond::generateDisassembly(Addr pc, return ss.str(); } -Addr +PowerISA::PCState BranchRegCond::branchTarget(ThreadContext *tc) const { uint32_t regVal = tc->readIntReg(_srcRegIdx[_numSrcRegs - 1]); - return (regVal & 0xfffffffc); + return regVal & 0xfffffffc; } std::string diff --git a/src/arch/power/insts/branch.hh b/src/arch/power/insts/branch.hh index dd00e42c3..7b9e78dee 100644 --- a/src/arch/power/insts/branch.hh +++ b/src/arch/power/insts/branch.hh @@ -86,7 +86,7 @@ class BranchPCRel : public PCDependentDisassembly } } - Addr branchTarget(Addr pc) const; + PowerISA::PCState branchTarget(const PowerISA::PCState &pc) const; std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const; }; @@ -112,7 +112,7 @@ class BranchNonPCRel : public PCDependentDisassembly } } - Addr branchTarget(Addr pc) const; + PowerISA::PCState branchTarget(const PowerISA::PCState &pc) const; std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const; }; @@ -187,7 +187,7 @@ class BranchPCRelCond : public BranchCond } } - Addr branchTarget(Addr pc) const; + PowerISA::PCState branchTarget(const PowerISA::PCState &pc) const; std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const; }; @@ -213,7 +213,7 @@ class BranchNonPCRelCond : public BranchCond } } - Addr branchTarget(Addr pc) const; + PowerISA::PCState branchTarget(const PowerISA::PCState &pc) const; std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const; }; @@ -231,7 +231,7 @@ class BranchRegCond : public BranchCond { } - Addr branchTarget(ThreadContext *tc) const; + PowerISA::PCState branchTarget(ThreadContext *tc) const; std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const; }; diff --git a/src/arch/power/insts/static_inst.hh b/src/arch/power/insts/static_inst.hh index 399e75371..91eca6fb0 100644 --- a/src/arch/power/insts/static_inst.hh +++ b/src/arch/power/insts/static_inst.hh @@ -63,6 +63,12 @@ class PowerStaticInst : public StaticInst std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const; + + void + advancePC(PowerISA::PCState &pcState) const + { + pcState.advance(); + } }; } // PowerISA namespace diff --git a/src/arch/power/isa/decoder.isa b/src/arch/power/isa/decoder.isa index 3252ff14a..671f57389 100644 --- a/src/arch/power/isa/decoder.isa +++ b/src/arch/power/isa/decoder.isa @@ -381,12 +381,20 @@ decode OPCODE default Unknown::unknown() { // Conditionally branch relative to PC based on CR and CTR. format BranchPCRelCondCtr { - 0: bc({{ NPC = PC + disp; }}); + 0: bc({{ + PowerISA::PCState pc = PCS; + pc.npc((uint32_t)(pc.pc() + disp)); + PCS = pc; + }}); } // Conditionally branch to fixed address based on CR and CTR. format BranchNonPCRelCondCtr { - 1: bca({{ NPC = targetAddr; }}); + 1: bca({{ + PowerISA::PCState pc = PCS; + pc.npc(targetAddr); + PCS = pc; + }}); } } @@ -394,12 +402,20 @@ decode OPCODE default Unknown::unknown() { // Unconditionally branch relative to PC. format BranchPCRel { - 0: b({{ NPC = PC + disp; }}); + 0: b({{ + PowerISA::PCState pc = PCS; + pc.npc((uint32_t)(pc.pc() + disp)); + PCS = pc; + }}); } // Unconditionally branch to fixed address. format BranchNonPCRel { - 1: ba({{ NPC = targetAddr; }}); + 1: ba({{ + PowerISA::PCState pc = PCS; + pc.npc(targetAddr); + PCS = pc; + }}); } } @@ -407,12 +423,20 @@ decode OPCODE default Unknown::unknown() { // Conditionally branch to address in LR based on CR and CTR. format BranchLrCondCtr { - 16: bclr({{ NPC = LR & 0xfffffffc; }}); + 16: bclr({{ + PowerISA::PCState pc = PCS; + pc.npc(LR & 0xfffffffc); + PCS = pc; + }}); } // Conditionally branch to address in CTR based on CR. format BranchCtrCond { - 528: bcctr({{ NPC = CTR & 0xfffffffc; }}); + 528: bcctr({{ + PowerISA::PCState pc = PCS; + pc.npc(CTR & 0xfffffffc); + PCS = pc; + }}); } // Condition register manipulation instructions. diff --git a/src/arch/power/isa/formats/branch.isa b/src/arch/power/isa/formats/branch.isa index d51ed5c25..da1579ea8 100644 --- a/src/arch/power/isa/formats/branch.isa +++ b/src/arch/power/isa/formats/branch.isa @@ -48,7 +48,7 @@ let {{ # Simple code to update link register (LR). -updateLrCode = 'LR = PC + 4;' +updateLrCode = 'PowerISA::PCState lrpc = PCS; LR = lrpc.pc() + 4;' }}; @@ -105,7 +105,7 @@ def GetCondCode(br_code): cond_code = 'if(condOk(CR)) {\n' cond_code += ' ' + br_code + '\n' cond_code += '} else {\n' - cond_code += ' NPC = NPC;\n' + cond_code += ' PCS = PCS;\n' cond_code += '}\n' return cond_code @@ -119,7 +119,7 @@ def GetCtrCondCode(br_code): cond_code += 'if(ctr_ok && cond_ok) {\n' cond_code += ' ' + br_code + '\n' cond_code += '} else {\n' - cond_code += ' NPC = NPC;\n' + cond_code += ' PCS = PCS;\n' cond_code += '}\n' cond_code += 'CTR = ctr;\n' return cond_code diff --git a/src/arch/power/isa/formats/unknown.isa b/src/arch/power/isa/formats/unknown.isa index 06e6ece26..8914cf9a6 100644 --- a/src/arch/power/isa/formats/unknown.isa +++ b/src/arch/power/isa/formats/unknown.isa @@ -76,7 +76,7 @@ output exec {{ { panic("attempt to execute unknown instruction at %#x" "(inst 0x%08x, opcode 0x%x, binary: %s)", - xc->readPC(), machInst, OPCODE, inst2string(machInst)); + xc->pcState().pc(), machInst, OPCODE, inst2string(machInst)); return new UnimplementedOpcodeFault; } }}; diff --git a/src/arch/power/isa/operands.isa b/src/arch/power/isa/operands.isa index fc6c32685..908e6e0e7 100644 --- a/src/arch/power/isa/operands.isa +++ b/src/arch/power/isa/operands.isa @@ -59,8 +59,7 @@ def operands {{ 'Mem': ('Mem', 'uw', None, ('IsMemRef', 'IsLoad', 'IsStore'), 8), # Program counter and next - 'PC': ('PC', 'uw', None, (None, None, 'IsControl'), 9), - 'NPC': ('NPC', 'uw', None, (None, None, 'IsControl'), 9), + 'PCS': ('PCState', 'uq', None, (None, None, 'IsControl'), 9), # Control registers 'CR': ('IntReg', 'uw', 'INTREG_CR', 'IsInteger', 9), diff --git a/src/arch/power/predecoder.hh b/src/arch/power/predecoder.hh index 1f3ac41cb..b1f2b6e38 100644 --- a/src/arch/power/predecoder.hh +++ b/src/arch/power/predecoder.hh @@ -83,7 +83,7 @@ class Predecoder // Use this to give data to the predecoder. This should be used // when there is control flow. void - moreBytes(Addr pc, Addr fetchPC, MachInst inst) + moreBytes(const PCState &pc, Addr fetchPC, MachInst inst) { emi = inst; } @@ -110,7 +110,7 @@ class Predecoder // This returns a constant reference to the ExtMachInst to avoid a copy const ExtMachInst & - getExtMachInst() + getExtMachInst(PCState &pcState) { return emi; } diff --git a/src/arch/power/process.cc b/src/arch/power/process.cc index 9fb69b9f8..a34a874bc 100644 --- a/src/arch/power/process.cc +++ b/src/arch/power/process.cc @@ -256,9 +256,7 @@ PowerLiveProcess::argsInit(int intSize, int pageSize) //Set the stack pointer register tc->setIntReg(StackPointerReg, stack_min); - Addr prog_entry = objFile->entryPoint(); - tc->setPC(prog_entry); - tc->setNextPC(prog_entry + sizeof(MachInst)); + tc->pcState(objFile->entryPoint()); //Align the "stack_min" to a page boundary. stack_min = roundDown(stack_min, pageSize); diff --git a/src/arch/power/types.hh b/src/arch/power/types.hh index 6a8d1e9d3..d049cdec1 100644 --- a/src/arch/power/types.hh +++ b/src/arch/power/types.hh @@ -31,6 +31,7 @@ #ifndef __ARCH_POWER_TYPES_HH__ #define __ARCH_POWER_TYPES_HH__ +#include "arch/generic/types.hh" #include "base/bitunion.hh" #include "base/hashmap.hh" #include "base/types.hh" @@ -78,6 +79,8 @@ BitUnion32(ExtMachInst) Bitfield<19, 12> fxm; EndBitUnion(ExtMachInst) +typedef GenericISA::SimplePCState PCState; + // typedef uint64_t LargestRead; // // Need to use 64 bits to make sure that read requests get handled properly diff --git a/src/arch/power/utility.cc b/src/arch/power/utility.cc index d48d4870a..399ec1f56 100644 --- a/src/arch/power/utility.cc +++ b/src/arch/power/utility.cc @@ -52,8 +52,7 @@ copyRegs(ThreadContext *src, ThreadContext *dest) copyMiscRegs(src, dest); // Lastly copy PC/NPC - dest->setPC(src->readPC()); - dest->setNextPC(src->readNextPC()); + dest->pcState(src->pcState()); } void diff --git a/src/arch/power/utility.hh b/src/arch/power/utility.hh index c8cd441ba..a47fcdc46 100644 --- a/src/arch/power/utility.hh +++ b/src/arch/power/utility.hh @@ -36,10 +36,19 @@ #define __ARCH_POWER_UTILITY_HH__ #include "base/types.hh" +#include "cpu/static_inst.hh" #include "cpu/thread_context.hh" namespace PowerISA { +inline PCState +buildRetPC(const PCState &curPC, const PCState &callPC) +{ + PCState retPC = callPC; + retPC.advance(); + return retPC; +} + /** * Function to ensure ISA semantics about 0 registers. * @param tc The thread context. @@ -63,6 +72,12 @@ copyMiscRegs(ThreadContext *src, ThreadContext *dest) void skipFunction(ThreadContext *tc); +inline void +advancePC(PCState &pc, const StaticInstPtr inst) +{ + pc.advance(); +} + } // PowerISA namespace #endif // __ARCH_POWER_UTILITY_HH__ diff --git a/src/arch/sparc/faults.cc b/src/arch/sparc/faults.cc index df0a283b9..28ee64321 100644 --- a/src/arch/sparc/faults.cc +++ b/src/arch/sparc/faults.cc @@ -307,15 +307,11 @@ void doREDFault(ThreadContext *tc, TrapType tt) //MiscReg CANSAVE = tc->readMiscRegNoEffect(MISCREG_CANSAVE); MiscReg CANSAVE = tc->readMiscRegNoEffect(NumIntArchRegs + 3); MiscReg GL = tc->readMiscRegNoEffect(MISCREG_GL); - MiscReg PC = tc->readPC(); - MiscReg NPC = tc->readNextPC(); + PCState pc = tc->pcState(); TL++; - if (bits(PSTATE, 3,3)) { - PC &= mask(32); - NPC &= mask(32); - } + Addr pcMask = bits(PSTATE, 3) ? mask(32) : mask(64); //set TSTATE.gl to gl replaceBits(TSTATE, 42, 40, GL); @@ -332,9 +328,9 @@ void doREDFault(ThreadContext *tc, TrapType tt) tc->setMiscRegNoEffect(MISCREG_TSTATE, TSTATE); //set TPC to PC - tc->setMiscRegNoEffect(MISCREG_TPC, PC); + tc->setMiscRegNoEffect(MISCREG_TPC, pc.pc() & pcMask); //set TNPC to NPC - tc->setMiscRegNoEffect(MISCREG_TNPC, NPC); + tc->setMiscRegNoEffect(MISCREG_TNPC, pc.npc() & pcMask); //set HTSTATE.hpstate to hpstate tc->setMiscRegNoEffect(MISCREG_HTSTATE, HPSTATE); @@ -394,18 +390,14 @@ void doNormalFault(ThreadContext *tc, TrapType tt, bool gotoHpriv) //MiscReg CANSAVE = tc->readMiscRegNoEffect(MISCREG_CANSAVE); MiscReg CANSAVE = tc->readIntReg(NumIntArchRegs + 3); MiscReg GL = tc->readMiscRegNoEffect(MISCREG_GL); - MiscReg PC = tc->readPC(); - MiscReg NPC = tc->readNextPC(); - - if (bits(PSTATE, 3,3)) { - PC &= mask(32); - NPC &= mask(32); - } + PCState pc = tc->pcState(); //Increment the trap level TL++; tc->setMiscRegNoEffect(MISCREG_TL, TL); + Addr pcMask = bits(PSTATE, 3) ? mask(32) : mask(64); + //Save off state //set TSTATE.gl to gl @@ -423,9 +415,9 @@ void doNormalFault(ThreadContext *tc, TrapType tt, bool gotoHpriv) tc->setMiscRegNoEffect(MISCREG_TSTATE, TSTATE); //set TPC to PC - tc->setMiscRegNoEffect(MISCREG_TPC, PC); + tc->setMiscRegNoEffect(MISCREG_TPC, pc.pc() & pcMask); //set TNPC to NPC - tc->setMiscRegNoEffect(MISCREG_TNPC, NPC); + tc->setMiscRegNoEffect(MISCREG_TNPC, pc.npc() & pcMask); //set HTSTATE.hpstate to hpstate tc->setMiscRegNoEffect(MISCREG_HTSTATE, HPSTATE); @@ -479,7 +471,7 @@ void doNormalFault(ThreadContext *tc, TrapType tt, bool gotoHpriv) } } -void getREDVector(MiscReg TT, Addr & PC, Addr & NPC) +void getREDVector(MiscReg TT, Addr &PC, Addr &NPC) { //XXX The following constant might belong in a header file. const Addr RSTVAddr = 0xFFF0000000ULL; @@ -554,9 +546,13 @@ void SparcFaultBase::invoke(ThreadContext * tc, StaticInstPtr inst) getPrivVector(tc, PC, NPC, trapType(), tl+1); } - tc->setPC(PC); - tc->setNextPC(NPC); - tc->setNextNPC(NPC + sizeof(MachInst)); + PCState pc; + pc.pc(PC); + pc.npc(NPC); + pc.nnpc(NPC + sizeof(MachInst)); + pc.upc(0); + pc.nupc(1); + tc->pcState(pc); } void PowerOnReset::invoke(ThreadContext * tc, StaticInstPtr inst) @@ -593,9 +589,14 @@ void PowerOnReset::invoke(ThreadContext * tc, StaticInstPtr inst) Addr PC, NPC; getREDVector(trapType(), PC, NPC); - tc->setPC(PC); - tc->setNextPC(NPC); - tc->setNextNPC(NPC + sizeof(MachInst)); + + PCState pc; + pc.pc(PC); + pc.npc(NPC); + pc.nnpc(NPC + sizeof(MachInst)); + pc.upc(0); + pc.nupc(1); + tc->pcState(pc); //These registers are specified as "undefined" after a POR, and they //should have reasonable values after the miscregfile is reset @@ -664,10 +665,7 @@ void SpillNNormal::invoke(ThreadContext *tc, StaticInstPtr inst) assert(lp); //Then adjust the PC and NPC - Addr spillStart = lp->readSpillStart(); - tc->setPC(spillStart); - tc->setNextPC(spillStart + sizeof(MachInst)); - tc->setNextNPC(spillStart + 2*sizeof(MachInst)); + tc->pcState(lp->readSpillStart()); } void FillNNormal::invoke(ThreadContext *tc, StaticInstPtr inst) @@ -681,10 +679,7 @@ void FillNNormal::invoke(ThreadContext *tc, StaticInstPtr inst) assert(lp); //Then adjust the PC and NPC - Addr fillStart = lp->readFillStart(); - tc->setPC(fillStart); - tc->setNextPC(fillStart + sizeof(MachInst)); - tc->setNextNPC(fillStart + 2*sizeof(MachInst)); + tc->pcState(lp->readFillStart()); } void TrapInstruction::invoke(ThreadContext *tc, StaticInstPtr inst) @@ -702,9 +697,9 @@ void TrapInstruction::invoke(ThreadContext *tc, StaticInstPtr inst) //We need to explicitly advance the pc, since that's not done for us //on a faulting instruction - tc->setPC(tc->readNextPC()); - tc->setNextPC(tc->readNextNPC()); - tc->setNextNPC(tc->readNextNPC() + sizeof(MachInst)); + PCState pc = tc->pcState(); + pc.advance(); + tc->pcState(pc); } #endif diff --git a/src/arch/sparc/isa/base.isa b/src/arch/sparc/isa/base.isa index ce063bcdc..c6055b3a3 100644 --- a/src/arch/sparc/isa/base.isa +++ b/src/arch/sparc/isa/base.isa @@ -111,6 +111,8 @@ output header {{ void printRegArray(std::ostream &os, const RegIndex indexArray[], int num) const; + + void advancePC(SparcISA::PCState &pcState) const; }; bool passesFpCondition(uint32_t fcc, uint32_t condition); @@ -260,6 +262,12 @@ output decoder {{ } } + void + SparcStaticInst::advancePC(SparcISA::PCState &pcState) const + { + pcState.advance(); + } + void SparcStaticInst::printSrcReg(std::ostream &os, int reg) const { diff --git a/src/arch/sparc/isa/decoder.isa b/src/arch/sparc/isa/decoder.isa index b9b38b569..80b29398c 100644 --- a/src/arch/sparc/isa/decoder.isa +++ b/src/arch/sparc/isa/decoder.isa @@ -46,14 +46,18 @@ decode OP default Unknown::unknown() { //Branch Always 0x8: bpa(19, annul_code={{ - NPC = xc->readPC() + disp; - NNPC = NPC + 4; + SparcISA::PCState pc = PCS; + pc.npc(pc.pc() + disp); + pc.nnpc(pc.npc() + 4); + PCS = pc; }}); //Branch Never 0x0: bpn(19, {{;}}, annul_code={{ - NNPC = NPC + 8; - NPC = NPC + 4; + SparcISA::PCState pc = PCS; + pc.nnpc(pc.npc() + 8); + pc.npc(pc.npc() + 4); + PCS = pc; }}); default: decode BPCC { @@ -66,14 +70,18 @@ decode OP default Unknown::unknown() { //Branch Always 0x8: ba(22, annul_code={{ - NPC = xc->readPC() + disp; - NNPC = NPC + 4; + SparcISA::PCState pc = PCS; + pc.npc(pc.pc() + disp); + pc.nnpc(pc.npc() + 4); + PCS = pc; }}); //Branch Never 0x0: bn(22, {{;}}, annul_code={{ - NNPC = NPC + 8; - NPC = NPC + 4; + SparcISA::PCState pc = PCS; + pc.nnpc(pc.npc() + 8); + pc.npc(pc.npc() + 4); + PCS = pc; }}); default: bicc(22, test={{passesCondition(Ccr<3:0>, COND2)}}); } @@ -97,14 +105,18 @@ decode OP default Unknown::unknown() format BranchN { //Branch Always 0x8: fbpa(22, annul_code={{ - NPC = xc->readPC() + disp; - NNPC = NPC + 4; + SparcISA::PCState pc = PCS; + pc.npc(pc.pc() + disp); + pc.nnpc(pc.npc() + 4); + PCS = pc; }}); //Branch Never 0x0: fbpn(22, {{;}}, annul_code={{ - NNPC = NPC + 8; - NPC = NPC + 4; + SparcISA::PCState pc = PCS; + pc.nnpc(pc.npc() + 8); + pc.npc(pc.npc() + 4); + PCS = pc; }}); default: decode BPCC { 0x0: fbpfcc0(19, test= @@ -123,14 +135,18 @@ decode OP default Unknown::unknown() format BranchN { //Branch Always 0x8: fba(22, annul_code={{ - NPC = xc->readPC() + disp; - NNPC = NPC + 4; + SparcISA::PCState pc = PCS; + pc.npc(pc.pc() + disp); + pc.nnpc(pc.npc() + 4); + PCS = pc; }}); //Branch Never 0x0: fbn(22, {{;}}, annul_code={{ - NNPC = NPC + 8; - NPC = NPC + 4; + SparcISA::PCState pc = PCS; + pc.nnpc(pc.npc() + 8); + pc.npc(pc.npc() + 4); + PCS = pc; }}); default: fbfcc(22, test= {{passesFpCondition(Fsr<11:10>, COND2)}}); @@ -138,11 +154,13 @@ decode OP default Unknown::unknown() } } 0x1: BranchN::call(30, {{ + SparcISA::PCState pc = PCS; if (Pstate<3:>) - R15 = (xc->readPC())<31:0>; + R15 = (pc.pc())<31:0>; else - R15 = xc->readPC(); - NNPC = R15 + disp; + R15 = pc.pc(); + pc.nnpc(R15 + disp); + PCS = pc; }}); 0x2: decode OP3 { format IntOp { @@ -316,10 +334,12 @@ decode OP default Unknown::unknown() 0x03: NoPriv::rdasi({{Rd = Asi;}}); 0x04: Priv::rdtick({{Rd = Tick;}}, {{Tick<63:>}}); 0x05: NoPriv::rdpc({{ + SparcISA::PCState pc = PCS; if(Pstate<3:>) - Rd = (xc->readPC())<31:0>; + Rd = (pc.pc())<31:0>; else - Rd = xc->readPC();}}); + Rd = pc.pc(); + }}); 0x06: NoPriv::rdfprs({{ //Wait for all fpops to finish. Rd = Fprs; @@ -973,7 +993,8 @@ decode OP default Unknown::unknown() 0x51: m5break({{PseudoInst::debugbreak(xc->tcBase()); }}, IsNonSpeculative); 0x54: m5panic({{ - panic("M5 panic instruction called at pc=%#x.", xc->readPC()); + SparcISA::PCState pc = PCS; + panic("M5 panic instruction called at pc=%#x.", pc.pc()); }}, No_OpClass, IsNonSpeculative); } #endif @@ -985,11 +1006,13 @@ decode OP default Unknown::unknown() fault = new MemAddressNotAligned; else { + SparcISA::PCState pc = PCS; if (Pstate<3:>) - Rd = (xc->readPC())<31:0>; + Rd = (pc.pc())<31:0>; else - Rd = xc->readPC(); - NNPC = target; + Rd = pc.pc(); + pc.nnpc(target); + PCS = pc; } }}); 0x39: Branch::return({{ @@ -1010,7 +1033,9 @@ decode OP default Unknown::unknown() fault = new MemAddressNotAligned; else { - NNPC = target; + SparcISA::PCState pc = PCS; + pc.nnpc(target); + PCS = pc; Cwp = (Cwp - 1 + NWindows) % NWindows; Cansave = Cansave + 1; Canrestore = Canrestore - 1; @@ -1082,8 +1107,10 @@ decode OP default Unknown::unknown() Ccr = Tstate<39:32>; Gl = Tstate<42:40>; Hpstate = Htstate; - NPC = Tnpc; - NNPC = Tnpc + 4; + SparcISA::PCState pc = PCS; + pc.npc(Tnpc); + pc.nnpc(Tnpc + 4); + PCS = pc; Tl = Tl - 1; }}, checkTl=true); 0x1: Priv::retry({{ @@ -1093,8 +1120,10 @@ decode OP default Unknown::unknown() Ccr = Tstate<39:32>; Gl = Tstate<42:40>; Hpstate = Htstate; - NPC = Tpc; - NNPC = Tnpc; + SparcISA::PCState pc = PCS; + pc.npc(Tpc); + pc.nnpc(Tnpc); + PCS = pc; Tl = Tl - 1; }}, checkTl=true); } diff --git a/src/arch/sparc/isa/formats/branch.isa b/src/arch/sparc/isa/formats/branch.isa index faaee8842..e62e0035a 100644 --- a/src/arch/sparc/isa/formats/branch.isa +++ b/src/arch/sparc/isa/formats/branch.isa @@ -193,7 +193,7 @@ def template JumpExecute {{ %(op_decl)s; %(op_rd)s; - NNPC = xc->readNextNPC(); + PCS = PCS; %(code)s; if(fault == NoFault) @@ -289,15 +289,24 @@ let {{ def doCondBranch(name, Name, base, cond, code, opt_flags): return doBranch(name, Name, base, cond, code, code, - 'NPC = NPC; NNPC = NNPC;', - 'NNPC = NPC + 8; NPC = NPC + 4', + 'PCS = PCS;', + ''' + SparcISA::PCState pc = PCS; + pc.nnpc(pc.npc() + 8); + pc.npc(pc.npc() + 4); + PCS = pc; + ''', opt_flags) def doUncondBranch(name, Name, base, code, annul_code, opt_flags): return doBranch(name, Name, base, "true", code, annul_code, ";", ";", opt_flags) - default_branch_code = "NNPC = xc->readPC() + disp;" + default_branch_code = ''' + SparcISA::PCState pc = PCS; + pc.nnpc(pc.pc() + disp); + PCS = pc; + ''' }}; // Format for branch instructions with n bit displacements: diff --git a/src/arch/sparc/isa/formats/micro.isa b/src/arch/sparc/isa/formats/micro.isa index c1d0c4f36..b5a53a68b 100644 --- a/src/arch/sparc/isa/formats/micro.isa +++ b/src/arch/sparc/isa/formats/micro.isa @@ -81,10 +81,11 @@ output header {{ StaticInstPtr * microops; - StaticInstPtr fetchMicroop(MicroPC microPC) + StaticInstPtr + fetchMicroop(MicroPC upc) const { - assert(microPC < numMicroops); - return microops[microPC]; + assert(upc < numMicroops); + return microops[upc]; } %(MacroExecute)s @@ -102,6 +103,15 @@ output header {{ { flags[IsMicroop] = true; } + + void + advancePC(SparcISA::PCState &pcState) const + { + if (flags[IsLastMicroop]) + pcState.uEnd(); + else + pcState.uAdvance(); + } }; class SparcDelayedMicroInst : public SparcMicroInst diff --git a/src/arch/sparc/isa/operands.isa b/src/arch/sparc/isa/operands.isa index a627a2e6f..8bf6450be 100644 --- a/src/arch/sparc/isa/operands.isa +++ b/src/arch/sparc/isa/operands.isa @@ -126,8 +126,7 @@ def operands {{ #'Frs2': ('FloatReg', 'df', 'dfpr(RS2)', 'IsFloating', 12), 'Frs2_low': ('FloatReg', 'uw', 'dfprl(RS2)', 'IsFloating', 12), 'Frs2_high': ('FloatReg', 'uw', 'dfprh(RS2)', 'IsFloating', 12), - 'NPC': ('NPC', 'udw', None, ( None, None, 'IsControl' ), 31), - 'NNPC': ('NNPC', 'udw', None, (None, None, 'IsControl' ), 32), + 'PCS': ('PCState', 'udw', None, (None, None, 'IsControl'), 30), # Registers which are used explicitly in instructions 'R0': ('IntReg', 'udw', '0', None, 6), 'R1': ('IntReg', 'udw', '1', None, 7), diff --git a/src/arch/sparc/nativetrace.cc b/src/arch/sparc/nativetrace.cc index 8a1eb7a58..9bccaaf7d 100644 --- a/src/arch/sparc/nativetrace.cc +++ b/src/arch/sparc/nativetrace.cc @@ -67,16 +67,17 @@ Trace::SparcNativeTrace::check(NativeTraceRecord *record) checkReg(*(regName++), regVal, realRegVal); } + SparcISA::PCState pc = tc->pcState(); // PC read(&realRegVal, sizeof(realRegVal)); realRegVal = SparcISA::gtoh(realRegVal); - regVal = tc->readNextPC(); + regVal = pc.npc(); checkReg("pc", regVal, realRegVal); // NPC read(&realRegVal, sizeof(realRegVal)); realRegVal = SparcISA::gtoh(realRegVal); - regVal = tc->readNextNPC(); + pc.nnpc(); checkReg("npc", regVal, realRegVal); // CCR diff --git a/src/arch/sparc/predecoder.hh b/src/arch/sparc/predecoder.hh index c4ab4fe79..8a2587929 100644 --- a/src/arch/sparc/predecoder.hh +++ b/src/arch/sparc/predecoder.hh @@ -72,7 +72,7 @@ namespace SparcISA //Use this to give data to the predecoder. This should be used //when there is control flow. - void moreBytes(Addr pc, Addr fetchPC, MachInst inst) + void moreBytes(const PCState &pc, Addr fetchPC, MachInst inst) { emi = inst; //The I bit, bit 13, is used to figure out where the ASI @@ -99,7 +99,7 @@ namespace SparcISA } //This returns a constant reference to the ExtMachInst to avoid a copy - const ExtMachInst & getExtMachInst() + const ExtMachInst & getExtMachInst(PCState &pcState) { return emi; } diff --git a/src/arch/sparc/process.cc b/src/arch/sparc/process.cc index 0cd8889a9..a6e21977a 100644 --- a/src/arch/sparc/process.cc +++ b/src/arch/sparc/process.cc @@ -69,41 +69,39 @@ SparcLiveProcess::SparcLiveProcess(LiveProcessParams * params, void SparcLiveProcess::handleTrap(int trapNum, ThreadContext *tc) { + PCState pc = tc->pcState(); switch(trapNum) { case 0x01: //Software breakpoint - warn("Software breakpoint encountered at pc %#x.\n", tc->readPC()); + warn("Software breakpoint encountered at pc %#x.\n", pc.pc()); break; case 0x02: //Division by zero - warn("Software signaled a division by zero at pc %#x.\n", - tc->readPC()); + warn("Software signaled a division by zero at pc %#x.\n", pc.pc()); break; case 0x03: //Flush window trap flushWindows(tc); break; case 0x04: //Clean windows warn("Ignoring process request for clean register " - "windows at pc %#x.\n", tc->readPC()); + "windows at pc %#x.\n", pc.pc()); break; case 0x05: //Range check - warn("Software signaled a range check at pc %#x.\n", - tc->readPC()); + warn("Software signaled a range check at pc %#x.\n", pc.pc()); break; case 0x06: //Fix alignment warn("Ignoring process request for os assisted unaligned accesses " - "at pc %#x.\n", tc->readPC()); + "at pc %#x.\n", pc.pc()); break; case 0x07: //Integer overflow - warn("Software signaled an integer overflow at pc %#x.\n", - tc->readPC()); + warn("Software signaled an integer overflow at pc %#x.\n", pc.pc()); break; case 0x32: //Get integer condition codes warn("Ignoring process request to get the integer condition codes " - "at pc %#x.\n", tc->readPC()); + "at pc %#x.\n", pc.pc()); break; case 0x33: //Set integer condition codes warn("Ignoring process request to set the integer condition codes " - "at pc %#x.\n", tc->readPC()); + "at pc %#x.\n", pc.pc()); break; default: panic("Unimplemented trap to operating system: trap number %#x.\n", trapNum); @@ -402,10 +400,7 @@ SparcLiveProcess::argsInit(int pageSize) // don't have anything like that, it should be set to 0. tc->setIntReg(1, 0); - Addr prog_entry = objFile->entryPoint(); - tc->setPC(prog_entry); - tc->setNextPC(prog_entry + sizeof(MachInst)); - tc->setNextNPC(prog_entry + (2 * sizeof(MachInst))); + tc->pcState(objFile->entryPoint()); //Align the "stack_min" to a page boundary. stack_min = roundDown(stack_min, pageSize); diff --git a/src/arch/sparc/remote_gdb.cc b/src/arch/sparc/remote_gdb.cc index 4eea0c077..aea40ea22 100644 --- a/src/arch/sparc/remote_gdb.cc +++ b/src/arch/sparc/remote_gdb.cc @@ -179,12 +179,14 @@ RemoteGDB::getregs() { memset(gdbregs.regs, 0, gdbregs.size); + PCState pc = context->pcState(); + if (context->readMiscReg(MISCREG_PSTATE) & PSTATE::am) { uint32_t *regs; regs = (uint32_t*)gdbregs.regs; - regs[Reg32Pc] = htobe((uint32_t)context->readPC()); - regs[Reg32Npc] = htobe((uint32_t)context->readNextPC()); + regs[Reg32Pc] = htobe((uint32_t)pc.pc()); + regs[Reg32Npc] = htobe((uint32_t)pc.npc()); for(int x = RegG0; x <= RegI0 + 7; x++) regs[x] = htobe((uint32_t)context->readIntReg(x - RegG0)); @@ -193,8 +195,8 @@ RemoteGDB::getregs() regs[Reg32Fsr] = htobe((uint32_t)context->readMiscReg(MISCREG_FSR)); regs[Reg32Csr] = htobe((uint32_t)context->readIntReg(NumIntArchRegs + 2)); } else { - gdbregs.regs[RegPc] = htobe(context->readPC()); - gdbregs.regs[RegNpc] = htobe(context->readNextPC()); + gdbregs.regs[RegPc] = htobe(pc.pc()); + gdbregs.regs[RegNpc] = htobe(pc.npc()); for(int x = RegG0; x <= RegI0 + 7; x++) gdbregs.regs[x] = htobe(context->readIntReg(x - RegG0)); @@ -224,8 +226,13 @@ RemoteGDB::getregs() void RemoteGDB::setregs() { - context->setPC(gdbregs.regs[RegPc]); - context->setNextPC(gdbregs.regs[RegNpc]); + PCState pc; + pc.pc(gdbregs.regs[RegPc]); + pc.npc(gdbregs.regs[RegNpc]); + pc.nnpc(pc.npc() + sizeof(MachInst)); + pc.upc(0); + pc.nupc(1); + context->pcState(pc); for(int x = RegG0; x <= RegI0 + 7; x++) context->setIntReg(x - RegG0, gdbregs.regs[x]); //Only the integer registers, pc and npc are set in netbsd @@ -241,6 +248,6 @@ RemoteGDB::clearSingleStep() void RemoteGDB::setSingleStep() { - nextBkpt = context->readNextPC(); + nextBkpt = context->pcState().npc(); setTempBreakpoint(nextBkpt); } diff --git a/src/arch/sparc/types.hh b/src/arch/sparc/types.hh index 70558ec6d..122fd808f 100644 --- a/src/arch/sparc/types.hh +++ b/src/arch/sparc/types.hh @@ -33,12 +33,15 @@ #include "base/bigint.hh" #include "base/types.hh" +#include "arch/generic/types.hh" namespace SparcISA { typedef uint32_t MachInst; typedef uint64_t ExtMachInst; + typedef GenericISA::DelaySlotUPCState PCState; + typedef Twin64_t LargestRead; struct CoreSpecific { diff --git a/src/arch/sparc/utility.cc b/src/arch/sparc/utility.cc index be4dfac84..6de44d8a3 100644 --- a/src/arch/sparc/utility.cc +++ b/src/arch/sparc/utility.cc @@ -213,20 +213,16 @@ copyRegs(ThreadContext *src, ThreadContext *dest) // Copy misc. registers copyMiscRegs(src, dest); - // Lastly copy PC/NPC - dest->setPC(src->readPC()); - dest->setNextPC(src->readNextPC()); - dest->setNextNPC(src->readNextNPC()); + dest->pcState(src->pcState()); } void skipFunction(ThreadContext *tc) { - Addr newpc = tc->readIntReg(ReturnAddressReg); - tc->setPC(newpc); - tc->setNextPC(tc->readPC() + sizeof(TheISA::MachInst)); - tc->setNextPC(tc->readNextPC() + sizeof(TheISA::MachInst)); + TheISA::PCState newPC = tc->pcState(); + newPC.set(tc->readIntReg(ReturnAddressReg)); + tc->pcState(newPC); } diff --git a/src/arch/sparc/utility.hh b/src/arch/sparc/utility.hh index f6a585e23..0d5a4bb93 100644 --- a/src/arch/sparc/utility.hh +++ b/src/arch/sparc/utility.hh @@ -36,11 +36,22 @@ #include "arch/sparc/tlb.hh" #include "base/misc.hh" #include "base/bitfield.hh" +#include "cpu/static_inst.hh" #include "cpu/thread_context.hh" #include "sim/fault.hh" namespace SparcISA { + + inline PCState + buildRetPC(const PCState &curPC, const PCState &callPC) + { + PCState ret = callPC; + ret.uEnd(); + ret.pc(curPC.npc()); + return ret; + } + uint64_t getArgument(ThreadContext *tc, int &number, uint16_t size, bool fp); @@ -78,6 +89,12 @@ namespace SparcISA void skipFunction(ThreadContext *tc); + inline void + advancePC(PCState &pc, const StaticInstPtr inst) + { + inst->advancePC(pc); + } + } // namespace SparcISA #endif diff --git a/src/arch/x86/faults.cc b/src/arch/x86/faults.cc index 4f2d97f90..7fb677c69 100644 --- a/src/arch/x86/faults.cc +++ b/src/arch/x86/faults.cc @@ -58,7 +58,8 @@ namespace X86ISA #if FULL_SYSTEM void X86FaultBase::invoke(ThreadContext * tc, StaticInstPtr inst) { - Addr pc = tc->readPC(); + PCState pcState = tc->pcState(); + Addr pc = pcState.pc(); DPRINTF(Faults, "RIP %#x: vector %d: %s\n", pc, vector, describe()); using namespace X86ISAInst::RomLabels; HandyM5Reg m5reg = tc->readMiscRegNoEffect(MISCREG_M5_REG); @@ -86,8 +87,9 @@ namespace X86ISA assert(!isSoft()); tc->setIntReg(INTREG_MICRO(15), errorCode); } - tc->setMicroPC(romMicroPC(entry)); - tc->setNextMicroPC(romMicroPC(entry) + 1); + pcState.upc(romMicroPC(entry)); + pcState.nupc(romMicroPC(entry) + 1); + tc->pcState(pcState); } std::string @@ -106,9 +108,8 @@ namespace X86ISA { X86FaultBase::invoke(tc); // This is the same as a fault, but it happens -after- the instruction. - tc->setPC(tc->readNextPC()); - tc->setNextPC(tc->readNextNPC()); - tc->setNextNPC(tc->readNextNPC() + sizeof(MachInst)); + PCState pc = tc->pcState(); + pc.uEnd(); } void X86Abort::invoke(ThreadContext * tc, StaticInstPtr inst) @@ -207,9 +208,8 @@ namespace X86ISA tc->setMiscReg(MISCREG_CS_LIMIT, 0xffffffff); tc->setMiscReg(MISCREG_CS_ATTR, codeAttr); - tc->setPC(0x000000000000fff0ULL + - tc->readMiscReg(MISCREG_CS_BASE)); - tc->setNextPC(tc->readPC() + sizeof(MachInst)); + PCState pc(0x000000000000fff0ULL + tc->readMiscReg(MISCREG_CS_BASE)); + tc->pcState(pc); tc->setMiscReg(MISCREG_TSG_BASE, 0); tc->setMiscReg(MISCREG_TSG_LIMIT, 0xffff); @@ -243,8 +243,9 @@ namespace X86ISA // Update the handy M5 Reg. tc->setMiscReg(MISCREG_M5_REG, 0); MicroPC entry = X86ISAInst::RomLabels::extern_label_initIntHalt; - tc->setMicroPC(romMicroPC(entry)); - tc->setNextMicroPC(romMicroPC(entry) + 1); + pc.upc(romMicroPC(entry)); + pc.nupc(romMicroPC(entry) + 1); + tc->pcState(pc); } void @@ -263,8 +264,7 @@ namespace X86ISA // This has the base value pre-added. tc->setMiscReg(MISCREG_CS_LIMIT, 0xffff); - tc->setPC(tc->readMiscReg(MISCREG_CS_BASE)); - tc->setNextPC(tc->readPC() + sizeof(MachInst)); + tc->pcState(tc->readMiscReg(MISCREG_CS_BASE)); } #else diff --git a/src/arch/x86/insts/macroop.hh b/src/arch/x86/insts/macroop.hh index 7ead7bdc2..fcf051a37 100644 --- a/src/arch/x86/insts/macroop.hh +++ b/src/arch/x86/insts/macroop.hh @@ -73,7 +73,8 @@ class MacroopBase : public X86StaticInst StaticInstPtr * microops; - StaticInstPtr fetchMicroop(MicroPC microPC) + StaticInstPtr + fetchMicroop(MicroPC microPC) const { assert(microPC < numMicroops); return microops[microPC]; diff --git a/src/arch/x86/insts/microop.hh b/src/arch/x86/insts/microop.hh index 9b0497efc..6fc215452 100644 --- a/src/arch/x86/insts/microop.hh +++ b/src/arch/x86/insts/microop.hh @@ -114,6 +114,15 @@ namespace X86ISA } bool checkCondition(uint64_t flags, int condition) const; + + void + advancePC(PCState &pcState) const + { + if (flags[IsLastMicroop]) + pcState.uEnd(); + else + pcState.uAdvance(); + } }; } diff --git a/src/arch/x86/insts/static_inst.hh b/src/arch/x86/insts/static_inst.hh index 2df5df092..8813f216c 100644 --- a/src/arch/x86/insts/static_inst.hh +++ b/src/arch/x86/insts/static_inst.hh @@ -158,6 +158,12 @@ namespace X86ISA panic("Tried to pick with unrecognized size %d.\n", size); } } + + void + advancePC(PCState &pcState) const + { + pcState.advance(); + } }; } diff --git a/src/arch/x86/isa/decoder/two_byte_opcodes.isa b/src/arch/x86/isa/decoder/two_byte_opcodes.isa index de167d1c1..def9b7f9d 100644 --- a/src/arch/x86/isa/decoder/two_byte_opcodes.isa +++ b/src/arch/x86/isa/decoder/two_byte_opcodes.isa @@ -199,7 +199,7 @@ #endif 0x54: m5panic({{ panic("M5 panic instruction called at pc=%#x.\n", - xc->readPC()); + xc->pcState().pc()); }}, IsNonSpeculative); 0x55: m5reserved1({{ warn("M5 reserved opcode 1 ignored.\n"); diff --git a/src/arch/x86/isa/formats/unknown.isa b/src/arch/x86/isa/formats/unknown.isa index 11751e861..1108fd4a4 100644 --- a/src/arch/x86/isa/formats/unknown.isa +++ b/src/arch/x86/isa/formats/unknown.isa @@ -47,13 +47,13 @@ output header {{ /** * Class for Unknown/Illegal instructions */ - class Unknown : public StaticInst + class Unknown : public X86ISA::X86StaticInst { public: // Constructor Unknown(ExtMachInst _machInst) : - StaticInst("unknown", _machInst, No_OpClass) + X86ISA::X86StaticInst("unknown", _machInst, No_OpClass) { } diff --git a/src/arch/x86/isa/microops/regop.isa b/src/arch/x86/isa/microops/regop.isa index 9ccea82dd..86ebac174 100644 --- a/src/arch/x86/isa/microops/regop.isa +++ b/src/arch/x86/isa/microops/regop.isa @@ -944,8 +944,12 @@ let {{ code = 'DoubleBits = psrc1 ^ op2;' class Wrip(WrRegOp, CondRegOp): - code = 'RIP = psrc1 + sop2 + CSBase' - else_code="RIP = RIP;" + code = ''' + X86ISA::PCState pc = PCS; + pc.npc(psrc1 + sop2 + CSBase); + PCS = pc; + ''' + else_code = "PCS = PCS;" class Wruflags(WrRegOp): code = 'ccFlagBits = psrc1 ^ op2' @@ -961,7 +965,10 @@ let {{ ''' class Rdip(RdRegOp): - code = 'DestReg = RIP - CSBase' + code = ''' + X86ISA::PCState pc = PCS; + DestReg = pc.npc() - CSBase; + ''' class Ruflags(RdRegOp): code = 'DestReg = ccFlagBits' diff --git a/src/arch/x86/isa/microops/seqop.isa b/src/arch/x86/isa/microops/seqop.isa index 57c44d48c..a3e22b0aa 100644 --- a/src/arch/x86/isa/microops/seqop.isa +++ b/src/arch/x86/isa/microops/seqop.isa @@ -169,15 +169,23 @@ let {{ return super(Eret, self).getAllocator(microFlags) iop = InstObjParams("br", "MicroBranchFlags", "SeqOpBase", - {"code": "nuIP = target", - "else_code": "nuIP = nuIP", + {"code": ''' + X86ISA::PCState pc = PCS; + pc.nupc(target); + PCS = pc; + ''', + "else_code": "PCS = PCS", "cond_test": "checkCondition(ccFlagBits, cc)"}) exec_output += SeqOpExecute.subst(iop) header_output += SeqOpDeclare.subst(iop) decoder_output += SeqOpConstructor.subst(iop) iop = InstObjParams("br", "MicroBranch", "SeqOpBase", - {"code": "nuIP = target", - "else_code": "nuIP = nuIP", + {"code": ''' + X86ISA::PCState pc = PCS; + pc.nupc(target); + PCS = pc; + ''', + "else_code": "PCS = PCS", "cond_test": "true"}) exec_output += SeqOpExecute.subst(iop) header_output += SeqOpDeclare.subst(iop) diff --git a/src/arch/x86/isa/operands.isa b/src/arch/x86/isa/operands.isa index d4140e414..25b73a8f2 100644 --- a/src/arch/x86/isa/operands.isa +++ b/src/arch/x86/isa/operands.isa @@ -97,9 +97,8 @@ def operands {{ 'FpSrcReg2': floatReg('src2', 21), 'FpDestReg': floatReg('dest', 22), 'FpData': floatReg('data', 23), - 'RIP': ('NPC', 'uqw', None, (None, None, 'IsControl'), 50), - 'uIP': ('UPC', 'uqw', None, (None, None, 'IsControl'), 51), - 'nuIP': ('NUPC', 'uqw', None, (None, None, 'IsControl'), 52), + 'PCS': ('PCState', 'udw', None, + (None, None, 'IsControl'), 50), # This holds the condition code portion of the flag register. The # nccFlagBits version holds the rest. 'ccFlagBits': intReg('INTREG_PSEUDO(0)', 60), diff --git a/src/arch/x86/nativetrace.cc b/src/arch/x86/nativetrace.cc index c5c891be9..6f92cfacf 100644 --- a/src/arch/x86/nativetrace.cc +++ b/src/arch/x86/nativetrace.cc @@ -85,7 +85,7 @@ X86NativeTrace::ThreadState::update(ThreadContext *tc) r13 = tc->readIntReg(X86ISA::INTREG_R13); r14 = tc->readIntReg(X86ISA::INTREG_R14); r15 = tc->readIntReg(X86ISA::INTREG_R15); - rip = tc->readNextPC(); + rip = tc->pcState().pc(); //This should be expanded if x87 registers are considered for (int i = 0; i < 8; i++) mmx[i] = tc->readFloatRegBits(X86ISA::FLOATREG_MMX(i)); diff --git a/src/arch/x86/predecoder.hh b/src/arch/x86/predecoder.hh index 5b38402e0..c06ec18bc 100644 --- a/src/arch/x86/predecoder.hh +++ b/src/arch/x86/predecoder.hh @@ -188,11 +188,11 @@ namespace X86ISA //Use this to give data to the predecoder. This should be used //when there is control flow. - void moreBytes(Addr pc, Addr fetchPC, MachInst data) + void moreBytes(const PCState &pc, Addr fetchPC, MachInst data) { DPRINTF(Predecoder, "Getting more bytes.\n"); basePC = fetchPC; - offset = (fetchPC >= pc) ? 0 : pc - fetchPC; + offset = (fetchPC >= pc.instAddr()) ? 0 : pc.instAddr() - fetchPC; fetchChunk = data; outOfBytes = false; process(); @@ -208,22 +208,26 @@ namespace X86ISA return emiIsReady; } + int + getInstSize() + { + int size = basePC + offset - origPC; + DPRINTF(Predecoder, + "Calculating the instruction size: " + "basePC: %#x offset: %#x origPC: %#x size: %d\n", + basePC, offset, origPC, size); + return size; + } + //This returns a constant reference to the ExtMachInst to avoid a copy - const ExtMachInst & getExtMachInst() + const ExtMachInst & + getExtMachInst(X86ISA::PCState &nextPC) { assert(emiIsReady); emiIsReady = false; + nextPC.npc(nextPC.pc() + getInstSize()); return emi; } - - int getInstSize() - { - DPRINTF(Predecoder, - "Calculating the instruction size: " - "basePC: %#x offset: %#x origPC: %#x\n", - basePC, offset, origPC); - return basePC + offset - origPC; - } }; }; diff --git a/src/arch/x86/process.cc b/src/arch/x86/process.cc index 946a7cbe1..bb875686e 100644 --- a/src/arch/x86/process.cc +++ b/src/arch/x86/process.cc @@ -116,10 +116,12 @@ X86_64LiveProcess::X86_64LiveProcess(LiveProcessParams *params, void I386LiveProcess::syscall(int64_t callnum, ThreadContext *tc) { - Addr eip = tc->readPC(); + TheISA::PCState pc = tc->pcState(); + Addr eip = pc.pc(); if (eip >= vsyscallPage.base && eip < vsyscallPage.base + vsyscallPage.size) { - tc->setNextPC(vsyscallPage.base + vsyscallPage.vsysexitOffset); + pc.npc(vsyscallPage.base + vsyscallPage.vsysexitOffset); + tc->pcState(pc); } X86LiveProcess::syscall(callnum, tc); } @@ -645,11 +647,9 @@ X86LiveProcess::argsInit(int pageSize, //Set the stack pointer register tc->setIntReg(StackPointerReg, stack_min); - Addr prog_entry = objFile->entryPoint(); // There doesn't need to be any segment base added in since we're dealing // with the flat segmentation model. - tc->setPC(prog_entry); - tc->setNextPC(prog_entry + sizeof(MachInst)); + tc->pcState(objFile->entryPoint()); //Align the "stack_min" to a page boundary. stack_min = roundDown(stack_min, pageSize); diff --git a/src/arch/x86/system.cc b/src/arch/x86/system.cc index ae47b14fd..3fc16e729 100644 --- a/src/arch/x86/system.cc +++ b/src/arch/x86/system.cc @@ -320,8 +320,7 @@ X86System::initState() cr0.pg = 1; tc->setMiscReg(MISCREG_CR0, cr0); - tc->setPC(tc->getSystemPtr()->kernelEntry); - tc->setNextPC(tc->readPC()); + tc->pcState(tc->getSystemPtr()->kernelEntry); // We should now be in long mode. Yay! diff --git a/src/arch/x86/tlb.cc b/src/arch/x86/tlb.cc index 71e0b3adb..dbba52af0 100644 --- a/src/arch/x86/tlb.cc +++ b/src/arch/x86/tlb.cc @@ -609,7 +609,7 @@ TLB::translate(RequestPtr req, ThreadContext *tc, Translation *translation, #else DPRINTF(TLB, "Handling a TLB miss for " "address %#x at pc %#x.\n", - vaddr, tc->readPC()); + vaddr, tc->instAddr()); Process *p = tc->getProcessPtr(); TlbEntry newEntry; diff --git a/src/arch/x86/types.hh b/src/arch/x86/types.hh index 2a0da7d65..5a208446a 100644 --- a/src/arch/x86/types.hh +++ b/src/arch/x86/types.hh @@ -42,6 +42,7 @@ #include +#include "arch/generic/types.hh" #include "base/bitunion.hh" #include "base/cprintf.hh" #include "base/hashmap.hh" @@ -221,6 +222,8 @@ namespace X86ISA return true; } + typedef GenericISA::UPCState PCState; + struct CoreSpecific { int core_type; }; diff --git a/src/arch/x86/utility.cc b/src/arch/x86/utility.cc index 624e8132f..88d5bfe58 100644 --- a/src/arch/x86/utility.cc +++ b/src/arch/x86/utility.cc @@ -72,8 +72,10 @@ void initCPU(ThreadContext *tc, int cpuId) InitInterrupt init(0); init.invoke(tc); - tc->setMicroPC(0); - tc->setNextMicroPC(1); + PCState pc = tc->pcState(); + pc.upc(0); + pc.nupc(1); + tc->pcState(pc); // These next two loops zero internal microcode and implicit registers. // They aren't specified by the ISA but are used internally by M5's @@ -231,8 +233,7 @@ copyRegs(ThreadContext *src, ThreadContext *dest) //copy float regs copyMiscRegs(src, dest); - dest->setPC(src->readPC()); - dest->setNextPC(src->readNextPC()); + dest->pcState(src->pcState()); } void diff --git a/src/arch/x86/utility.hh b/src/arch/x86/utility.hh index 05ce53347..143fde00c 100644 --- a/src/arch/x86/utility.hh +++ b/src/arch/x86/utility.hh @@ -46,12 +46,22 @@ #include "base/misc.hh" #include "base/types.hh" #include "config/full_system.hh" +#include "cpu/static_inst.hh" #include "cpu/thread_context.hh" class ThreadContext; namespace X86ISA { + + inline PCState + buildRetPC(const PCState &curPC, const PCState &callPC) + { + PCState retPC = callPC; + retPC.uEnd(); + return retPC; + } + uint64_t getArgument(ThreadContext *tc, int &number, uint16_t size, bool fp); @@ -86,6 +96,12 @@ namespace X86ISA void copyMiscRegs(ThreadContext *src, ThreadContext *dest); void skipFunction(ThreadContext *tc); + + inline void + advancePC(PCState &pc, const StaticInstPtr inst) + { + inst->advancePC(pc); + } }; #endif // __ARCH_X86_UTILITY_HH__ diff --git a/src/base/remote_gdb.cc b/src/base/remote_gdb.cc index 68747b3d1..4a06ff307 100644 --- a/src/base/remote_gdb.cc +++ b/src/base/remote_gdb.cc @@ -645,8 +645,8 @@ BaseRemoteGDB::trap(int type) bufferSize = gdbregs.bytes() * 2 + 256; buffer = (char*)malloc(bufferSize); - DPRINTF(GDBMisc, "trap: PC=%#x NPC=%#x\n", - context->readPC(), context->readNextPC()); + TheISA::PCState pc = context->pcState(); + DPRINTF(GDBMisc, "trap: PC=%s\n", pc); clearSingleStep(); @@ -806,8 +806,7 @@ BaseRemoteGDB::trap(int type) subcmd = hex2i(&p); if (*p++ == ';') { val = hex2i(&p); - context->setPC(val); - context->setNextPC(val + sizeof(MachInst)); + context->pcState(val); } clearSingleStep(); goto out; @@ -815,8 +814,7 @@ BaseRemoteGDB::trap(int type) case GDBCont: if (p - data < (ptrdiff_t)datalen) { val = hex2i(&p); - context->setPC(val); - context->setNextPC(val + sizeof(MachInst)); + context->pcState(val); } clearSingleStep(); goto out; @@ -825,8 +823,7 @@ BaseRemoteGDB::trap(int type) subcmd = hex2i(&p); if (*p++ == ';') { val = hex2i(&p); - context->setPC(val); - context->setNextPC(val + sizeof(MachInst)); + context->pcState(val); } setSingleStep(); goto out; @@ -834,8 +831,7 @@ BaseRemoteGDB::trap(int type) case GDBStep: if (p - data < (ptrdiff_t)datalen) { val = hex2i(&p); - context->setPC(val); - context->setNextPC(val + sizeof(MachInst)); + context->pcState(val); } setSingleStep(); goto out; diff --git a/src/base/types.hh b/src/base/types.hh index 30b2d9258..5ce778572 100644 --- a/src/base/types.hh +++ b/src/base/types.hh @@ -67,6 +67,28 @@ const Tick MaxTick = LL(0x7fffffffffffffff); */ typedef uint64_t Addr; +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; +} + const Addr MaxAddr = (Addr)-1; /** 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 #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::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::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 BaseDynInst::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::BaseDynInst(StaticInstPtr _staticInst, template BaseDynInst::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 void BaseDynInst::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::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 @@ -987,47 +987,6 @@ InOrderCPU::getPipeStage(int stage_num) return pipelineStage[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) { @@ -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 #include +#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 (cache_pkt->getPtr())); 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(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 @@ -233,30 +233,6 @@ InOrderThreadContext::setRegOtherThread(int misc_reg, const MiscReg &val, cpu->setRegOtherThread(misc_reg, val, tid); } -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) { 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, "", 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 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 #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::takeOverFrom() template bool -BPredUnit::predict(DynInstPtr &inst, Addr &PC, ThreadID tid) +BPredUnit::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::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::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::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::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::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::squash(const InstSeqNum &squashed_sn, ThreadID tid) template void BPredUnit::squash(const InstSeqNum &squashed_sn, - const Addr &corr_target, + const TheISA::PCState &corrTarget, bool actually_taken, ThreadID tid) { @@ -330,8 +327,8 @@ BPredUnit::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::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::BPSquash(void *bp_history) template bool -BPredUnit::BPLookup(Addr &inst_PC, void * &bp_history) +BPredUnit::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::BPLookup(Addr &inst_PC, void * &bp_history) template void -BPredUnit::BPUpdate(Addr &inst_PC, bool taken, void *bp_history) +BPredUnit::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::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 +#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::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::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 @@ -524,7 +518,7 @@ DefaultCommit::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::setArchFloatRegInt(int reg_idx, uint64_t val, ThreadID tid) } template -uint64_t -FullO3CPU::readPC(ThreadID tid) +TheISA::PCState +FullO3CPU::pcState(ThreadID tid) { - return commit.readPC(tid); + return commit.pcState(tid); } template void -FullO3CPU::setPC(Addr new_PC, ThreadID tid) -{ - commit.setPC(new_PC, tid); -} - -template -uint64_t -FullO3CPU::readMicroPC(ThreadID tid) +FullO3CPU::pcState(const TheISA::PCState &val, ThreadID tid) { - return commit.readMicroPC(tid); + commit.pcState(val, tid); } template -void -FullO3CPU::setMicroPC(Addr new_PC, ThreadID tid) +Addr +FullO3CPU::instAddr(ThreadID tid) { - commit.setMicroPC(new_PC, tid); + return commit.instAddr(tid); } template -uint64_t -FullO3CPU::readNextPC(ThreadID tid) +Addr +FullO3CPU::nextInstAddr(ThreadID tid) { - return commit.readNextPC(tid); + return commit.nextInstAddr(tid); } template -void -FullO3CPU::setNextPC(uint64_t val, ThreadID tid) -{ - commit.setNextPC(val, tid); -} - -template -uint64_t -FullO3CPU::readNextNPC(ThreadID tid) -{ - return commit.readNextNPC(tid); -} - -template -void -FullO3CPU::setNextNPC(uint64_t val, ThreadID tid) -{ - commit.setNextNPC(val, tid); -} - -template -uint64_t -FullO3CPU::readNextMicroPC(ThreadID tid) -{ - return commit.readNextMicroPC(tid); -} - -template -void -FullO3CPU::setNextMicroPC(Addr new_PC, ThreadID tid) +MicroPC +FullO3CPU::microPC(ThreadID tid) { - commit.setNextMicroPC(new_PC, tid); + return commit.microPC(tid); } template @@ -1419,9 +1382,9 @@ template void FullO3CPU::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::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::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::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 void DefaultDecode::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::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::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::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::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::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::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 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 BaseO3DynInst::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(staticInst, PC, NPC, microPC, - Pred_PC, Pred_NPC, Pred_MicroPC, seq_num, cpu) + : BaseDynInst(staticInst, pc, predPC, seq_num, cpu) { initVars(); } template BaseO3DynInst::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(inst, PC, NPC, microPC, - Pred_PC, Pred_NPC, Pred_MicroPC, seq_num, cpu) + : BaseDynInst(inst, pc, predPC, seq_num, cpu) { initVars(); } @@ -131,15 +125,17 @@ BaseO3DynInst::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::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::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::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::switchToInactive() template bool -DefaultFetch::lookupAndUpdateNextPC(DynInstPtr &inst, Addr &next_PC, - Addr &next_NPC, Addr &next_MicroPC) +DefaultFetch::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::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::fetchCacheLine(Addr fetch_PC, Fault &ret_fault, ThreadID tid template inline void -DefaultFetch::doSquash(const Addr &new_PC, - const Addr &new_NPC, const Addr &new_microPC, ThreadID tid) +DefaultFetch::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::doSquash(const Addr &new_PC, template void -DefaultFetch::squashFromDecode(const Addr &new_PC, const Addr &new_NPC, - const Addr &new_MicroPC, +DefaultFetch::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::updateFetchStatus() template void -DefaultFetch::squash(const Addr &new_PC, const Addr &new_NPC, - const Addr &new_MicroPC, +DefaultFetch::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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 void DefaultIEW::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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 @@ -981,8 +981,8 @@ InstructionQueue::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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::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 void MemDepUnit::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::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 @@ -518,9 +516,9 @@ void MemDepUnit::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 @@ -559,9 +557,9 @@ MemDepUnit::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::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::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::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::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::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::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::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::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::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::setFloatRegBits(int reg_idx, FloatRegBits val) template void -O3ThreadContext::setPC(uint64_t val) +O3ThreadContext::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 -void -O3ThreadContext::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 -void -O3ThreadContext::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 -void -O3ThreadContext::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 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 +#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 addrStack; + std::vector 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 #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(); } diff --git a/src/kern/system_events.cc b/src/kern/system_events.cc index 25856e466..612aabc0b 100644 --- a/src/kern/system_events.cc +++ b/src/kern/system_events.cc @@ -41,9 +41,10 @@ using namespace TheISA; void SkipFuncEvent::process(ThreadContext *tc) { - DPRINTF(PCEvent, "skipping %s: pc=%x, newpc=%x\n", description, - tc->readPC(), tc->readIntReg(ReturnAddressReg)); + TheISA::PCState oldPC M5_VAR_USED = tc->pcState(); // Call ISA specific code to do the skipping TheISA::skipFunction(tc); + DPRINTF(PCEvent, "skipping %s: pc = %s, newpc = %s\n", description, + oldPC, tc->pcState()); } diff --git a/src/kern/tru64/tru64.hh b/src/kern/tru64/tru64.hh index d3348fe89..aa3c4bb50 100644 --- a/src/kern/tru64/tru64.hh +++ b/src/kern/tru64/tru64.hh @@ -509,7 +509,9 @@ class Tru64 : public OperatingSystem // Note that we'll advance PC <- NPC before the end of the cycle, // so we need to restore the desired PC into NPC. // The current regs->pc will get clobbered. - tc->setNextPC(htog(sc->sc_pc)); + PCState pc = tc->pcState(); + pc.npc(htog(sc->sc_pc)); + tc->pcState(pc); for (int i = 0; i < 31; ++i) { tc->setIntReg(i, htog(sc->sc_regs[i])); @@ -703,8 +705,7 @@ class Tru64 : public OperatingSystem tc->setIntReg(TheISA::StackPointerReg, gtoh(attrp->registers.sp)); tc->setMiscRegNoEffect(AlphaISA::MISCREG_UNIQ, uniq_val); - tc->setPC(gtoh(attrp->registers.pc)); - tc->setNextPC(gtoh(attrp->registers.pc) + sizeof(TheISA::MachInst)); + tc->pcState(gtoh(attrp->registers.pc)); tc->activate(); } diff --git a/src/sim/faults.cc b/src/sim/faults.cc index 78b9fb0a4..fbe8fb32a 100644 --- a/src/sim/faults.cc +++ b/src/sim/faults.cc @@ -40,13 +40,12 @@ #if !FULL_SYSTEM void FaultBase::invoke(ThreadContext * tc, StaticInstPtr inst) { - panic("fault (%s) detected @ PC %p", name(), tc->readPC()); + panic("fault (%s) detected @ PC %s", name(), tc->pcState()); } #else void FaultBase::invoke(ThreadContext * tc, StaticInstPtr inst) { - DPRINTF(Fault, "Fault %s at PC: %#x\n", name(), tc->readPC()); - + DPRINTF(Fault, "Fault %s at PC: %s\n", name(), tc->pcState()); assert(!tc->misspeculating()); } #endif diff --git a/src/sim/insttracer.hh b/src/sim/insttracer.hh index 845f4ed99..1ff67d2cb 100644 --- a/src/sim/insttracer.hh +++ b/src/sim/insttracer.hh @@ -54,9 +54,8 @@ class InstRecord // need to make this ref-counted so it doesn't go away before we // dump the record StaticInstPtr staticInst; - Addr PC; + TheISA::PCState pc; StaticInstPtr macroStaticInst; - MicroPC upc; bool misspeculating; bool predicate; @@ -90,12 +89,11 @@ class InstRecord public: InstRecord(Tick _when, ThreadContext *_thread, const StaticInstPtr _staticInst, - Addr _pc, bool spec, - const StaticInstPtr _macroStaticInst = NULL, - MicroPC _upc = 0) + TheISA::PCState _pc, bool spec, + const StaticInstPtr _macroStaticInst = NULL) : when(_when), thread(_thread), - staticInst(_staticInst), PC(_pc), - macroStaticInst(_macroStaticInst), upc(_upc), + staticInst(_staticInst), pc(_pc), + macroStaticInst(_macroStaticInst), misspeculating(spec), predicate(true) { data_status = DataInvalid; @@ -137,9 +135,8 @@ class InstRecord Tick getWhen() { return when; } ThreadContext *getThread() { return thread; } StaticInstPtr getStaticInst() { return staticInst; } - Addr getPC() { return PC; } + TheISA::PCState getPCState() { return pc; } StaticInstPtr getMacroStaticInst() { return macroStaticInst; } - MicroPC getUPC() { return upc; } bool getMisspeculating() { return misspeculating; } Addr getAddr() { return addr; } @@ -167,9 +164,8 @@ class InstTracer : public SimObject virtual InstRecord * getInstRecord(Tick when, ThreadContext *tc, - const StaticInstPtr staticInst, Addr pc, - const StaticInstPtr macroStaticInst = NULL, - MicroPC _upc = 0) = 0; + const StaticInstPtr staticInst, TheISA::PCState pc, + const StaticInstPtr macroStaticInst = NULL) = 0; }; diff --git a/src/sim/syscall_emul.cc b/src/sim/syscall_emul.cc index 207c07309..60bb59790 100644 --- a/src/sim/syscall_emul.cc +++ b/src/sim/syscall_emul.cc @@ -824,9 +824,7 @@ cloneFunc(SyscallDesc *desc, int callnum, LiveProcess *process, ctc->setIntReg(TheISA::SyscallPseudoReturnReg, 1); #endif - ctc->setPC(tc->readNextPC()); - ctc->setNextPC(tc->readNextPC() + sizeof(TheISA::MachInst)); - ctc->setNextNPC(tc->readNextNPC() + sizeof(TheISA::MachInst)); + ctc->pcState(tc->nextInstAddr()); ctc->activate(); diff --git a/src/sim/syscall_emul.hh b/src/sim/syscall_emul.hh index eaec57ef5..f2847b2e5 100644 --- a/src/sim/syscall_emul.hh +++ b/src/sim/syscall_emul.hh @@ -512,8 +512,8 @@ ioctlFunc(SyscallDesc *desc, int callnum, LiveProcess *process, return -ENOTTY; default: - fatal("Unsupported ioctl call: ioctl(%d, 0x%x, ...) @ 0x%llx\n", - fd, req, tc->readPC()); + fatal("Unsupported ioctl call: ioctl(%d, 0x%x, ...) @ \n", + fd, req, tc->pcState()); } } -- 2.30.2