From 47877cf2dbd6ee2f1cf9b2c609d37b0589e876ca Mon Sep 17 00:00:00 2001 From: Nathan Binkert Date: Tue, 26 May 2009 09:23:13 -0700 Subject: [PATCH] types: add a type for thread IDs and try to use it everywhere --- src/arch/alpha/miscregfile.cc | 8 +- src/arch/alpha/miscregfile.hh | 13 +- src/arch/mips/mt.hh | 6 +- src/arch/mips/regfile.cc | 19 +- src/arch/mips/regfile/float_regfile.cc | 13 +- src/arch/mips/regfile/float_regfile.hh | 10 +- src/arch/mips/regfile/misc_regfile.cc | 27 ++- src/arch/mips/regfile/misc_regfile.hh | 18 +- src/arch/mips/regfile/regfile.cc | 12 +- src/arch/mips/regfile/regfile.hh | 12 +- src/base/types.hh | 6 + src/cpu/base.cc | 63 +++--- src/cpu/base.hh | 4 +- src/cpu/base_dyn_inst.hh | 4 +- src/cpu/btb.cc | 6 +- src/cpu/inorder/cpu.cc | 193 ++++++++-------- src/cpu/inorder/cpu.hh | 138 ++++++------ src/cpu/inorder/first_stage.cc | 36 +-- src/cpu/inorder/first_stage.hh | 14 +- src/cpu/inorder/inorder_cpu_builder.cc | 2 +- src/cpu/inorder/inorder_dyn_inst.cc | 11 +- src/cpu/inorder/inorder_dyn_inst.hh | 14 +- src/cpu/inorder/pipeline_stage.cc | 84 ++++--- src/cpu/inorder/pipeline_stage.hh | 40 ++-- src/cpu/inorder/resource.cc | 5 +- src/cpu/inorder/resource.hh | 15 +- src/cpu/inorder/resource_pool.9stage.cc | 14 +- src/cpu/inorder/resource_pool.cc | 16 +- src/cpu/inorder/resource_pool.hh | 19 +- src/cpu/inorder/resources/agen_unit.cc | 22 +- src/cpu/inorder/resources/bpred_unit.cc | 10 +- src/cpu/inorder/resources/bpred_unit.hh | 14 +- src/cpu/inorder/resources/branch_predictor.cc | 6 +- src/cpu/inorder/resources/branch_predictor.hh | 4 +- src/cpu/inorder/resources/cache_unit.cc | 46 ++-- src/cpu/inorder/resources/cache_unit.hh | 2 +- src/cpu/inorder/resources/decode_unit.cc | 24 +- src/cpu/inorder/resources/decode_unit.hh | 3 +- src/cpu/inorder/resources/execution_unit.cc | 4 +- src/cpu/inorder/resources/fetch_seq_unit.cc | 14 +- src/cpu/inorder/resources/fetch_seq_unit.hh | 8 +- src/cpu/inorder/resources/graduation_unit.cc | 20 +- src/cpu/inorder/resources/inst_buffer.cc | 18 +- src/cpu/inorder/resources/inst_buffer.hh | 6 +- src/cpu/inorder/resources/inst_buffer_new.cc | 4 +- src/cpu/inorder/resources/inst_buffer_new.hh | 2 +- src/cpu/inorder/resources/mult_div_unit.cc | 4 +- src/cpu/inorder/resources/tlb_unit.cc | 12 +- src/cpu/inorder/resources/tlb_unit.hh | 3 +- src/cpu/inorder/resources/use_def.cc | 7 +- src/cpu/inorder/resources/use_def.hh | 3 +- src/cpu/inorder/thread_context.cc | 5 +- src/cpu/inorder/thread_context.hh | 5 +- src/cpu/inorder/thread_state.hh | 6 +- src/cpu/o3/bpred_unit.hh | 14 +- src/cpu/o3/bpred_unit_impl.hh | 10 +- src/cpu/o3/btb.hh | 8 +- src/cpu/o3/commit.hh | 52 ++--- src/cpu/o3/commit_impl.hh | 177 +++++++-------- src/cpu/o3/cpu.cc | 208 ++++++++++-------- src/cpu/o3/cpu.hh | 129 ++++++----- src/cpu/o3/cpu_builder.cc | 4 +- src/cpu/o3/decode.hh | 26 +-- src/cpu/o3/decode_impl.hh | 77 +++---- src/cpu/o3/fetch.hh | 39 ++-- src/cpu/o3/fetch_impl.hh | 120 +++++----- src/cpu/o3/free_list.cc | 2 +- src/cpu/o3/free_list.hh | 2 +- src/cpu/o3/iew.hh | 37 ++-- src/cpu/o3/iew_impl.hh | 131 +++++------ src/cpu/o3/inst_queue.hh | 20 +- src/cpu/o3/inst_queue_impl.hh | 81 ++++--- src/cpu/o3/lsq.hh | 76 +++---- src/cpu/o3/lsq_impl.hh | 133 +++++------ src/cpu/o3/lsq_unit.hh | 2 +- src/cpu/o3/lsq_unit_impl.hh | 2 +- src/cpu/o3/mem_dep_unit.hh | 4 +- src/cpu/o3/mem_dep_unit_impl.hh | 16 +- src/cpu/o3/regfile.hh | 23 +- src/cpu/o3/rename.hh | 42 ++-- src/cpu/o3/rename_impl.hh | 136 ++++++------ src/cpu/o3/rob.hh | 46 ++-- src/cpu/o3/rob_impl.hh | 92 ++++---- src/cpu/o3/store_set.cc | 5 +- src/cpu/o3/store_set.hh | 5 +- src/cpu/o3/thread_context_impl.hh | 2 +- src/cpu/ozone/back_end_impl.hh | 88 ++++---- src/cpu/ozone/cpu_builder.cc | 4 +- src/cpu/ozone/inst_queue.hh | 14 +- src/cpu/ozone/inst_queue_impl.hh | 60 ++--- src/cpu/ozone/lw_back_end_impl.hh | 76 +++---- src/cpu/ozone/null_predictor.hh | 8 +- src/cpu/ozone/simple_cpu_builder.cc | 4 +- src/cpu/simple/atomic.cc | 6 +- src/cpu/simple/base.cc | 2 +- src/cpu/simple/base.hh | 5 +- src/cpu/simple/timing.cc | 8 +- src/cpu/simple_thread.hh | 12 +- src/cpu/smt.hh | 4 +- src/cpu/thread_context.hh | 11 +- src/cpu/thread_state.cc | 4 +- src/cpu/thread_state.hh | 10 +- src/mem/cache/mshr.cc | 2 +- src/mem/cache/mshr.hh | 2 +- src/mem/request.hh | 6 +- src/mem/ruby/system/CacheMemory.hh | 6 +- src/sim/system.hh | 2 +- 107 files changed, 1595 insertions(+), 1534 deletions(-) diff --git a/src/arch/alpha/miscregfile.cc b/src/arch/alpha/miscregfile.cc index 5dc316a83..d52d900fd 100644 --- a/src/arch/alpha/miscregfile.cc +++ b/src/arch/alpha/miscregfile.cc @@ -65,7 +65,7 @@ MiscRegFile::MiscRegFile(BaseCPU *_cpu) MiscReg -MiscRegFile::readRegNoEffect(int misc_reg, unsigned tid ) +MiscRegFile::readRegNoEffect(int misc_reg, ThreadID tid) { switch (misc_reg) { case MISCREG_FPCR: @@ -85,7 +85,7 @@ MiscRegFile::readRegNoEffect(int misc_reg, unsigned tid ) } MiscReg -MiscRegFile::readReg(int misc_reg, ThreadContext *tc, unsigned tid ) +MiscRegFile::readReg(int misc_reg, ThreadContext *tc, ThreadID tid) { switch (misc_reg) { case MISCREG_FPCR: @@ -104,7 +104,7 @@ MiscRegFile::readReg(int misc_reg, ThreadContext *tc, unsigned tid ) } void -MiscRegFile::setRegNoEffect(int misc_reg, const MiscReg &val, unsigned tid) +MiscRegFile::setRegNoEffect(int misc_reg, const MiscReg &val, ThreadID tid) { switch (misc_reg) { case MISCREG_FPCR: @@ -131,7 +131,7 @@ MiscRegFile::setRegNoEffect(int misc_reg, const MiscReg &val, unsigned tid) void MiscRegFile::setReg(int misc_reg, const MiscReg &val, ThreadContext *tc, - unsigned tid) + ThreadID tid) { switch (misc_reg) { case MISCREG_FPCR: diff --git a/src/arch/alpha/miscregfile.hh b/src/arch/alpha/miscregfile.hh index 1a215b8e4..0d95e5a6d 100644 --- a/src/arch/alpha/miscregfile.hh +++ b/src/arch/alpha/miscregfile.hh @@ -88,11 +88,12 @@ class MiscRegFile int getInstAsid(); int getDataAsid(); - MiscReg readRegNoEffect(int misc_reg, unsigned tid = 0); - MiscReg readReg(int misc_reg, ThreadContext *tc, unsigned tid = 0); + MiscReg readRegNoEffect(int misc_reg, ThreadID tid = 0); + MiscReg readReg(int misc_reg, ThreadContext *tc, ThreadID tid = 0); - void setRegNoEffect(int misc_reg, const MiscReg &val, unsigned tid = 0); - void setReg(int misc_reg, const MiscReg &val, ThreadContext *tc, unsigned tid = 0); + void setRegNoEffect(int misc_reg, const MiscReg &val, ThreadID tid = 0); + void setReg(int misc_reg, const MiscReg &val, ThreadContext *tc, + ThreadID tid = 0); void clear() @@ -107,12 +108,12 @@ class MiscRegFile void serialize(std::ostream &os); void unserialize(Checkpoint *cp, const std::string §ion); - void reset(std::string core_name, unsigned num_threads, + void reset(std::string core_name, ThreadID num_threads, unsigned num_vpes, BaseCPU *_cpu) { } - void expandForMultithreading(unsigned num_threads, unsigned num_vpes) + void expandForMultithreading(ThreadID num_threads, unsigned num_vpes) { } diff --git a/src/arch/mips/mt.hh b/src/arch/mips/mt.hh index d0c333d86..8b4c9f908 100755 --- a/src/arch/mips/mt.hh +++ b/src/arch/mips/mt.hh @@ -109,7 +109,7 @@ forkThread(TC *tc, Fault &fault, int Rd_bits, int Rs, int Rt) int num_threads = bits(tc->readMiscRegNoEffect(MVPConf0), MVPC0_PTC_HI, MVPC0_PTC_LO) + 1; int success = 0; - for (int tid = 0; tid < num_threads && success == 0; tid++) { + for (ThreadID tid = 0; tid < num_threads && success == 0; tid++) { unsigned tid_TCBind = tc->readRegOtherThread(MipsISA::TCBind + Ctrl_Base_DepTag, tid); unsigned tc_bind = tc->readMiscRegNoEffect(MipsISA::TCBind); @@ -181,7 +181,7 @@ yieldThread(TC *tc, Fault &fault, int src_reg, uint32_t yield_mask) { if (src_reg == 0) { unsigned mvpconf0 = tc->readMiscRegNoEffect(MVPConf0); - int num_threads = bits(mvpconf0, MVPC0_PTC_HI, MVPC0_PTC_LO) + 1; + ThreadID num_threads = bits(mvpconf0, MVPC0_PTC_HI, MVPC0_PTC_LO) + 1; int ok = 0; @@ -190,7 +190,7 @@ yieldThread(TC *tc, Fault &fault, int src_reg, uint32_t yield_mask) unsigned cur_vpe = bits(tcbind, TCB_CUR_VPE_HI, TCB_CUR_VPE_LO); unsigned cur_tc = bits(tcbind, TCB_CUR_TC_HI, TCB_CUR_TC_LO); - for (int tid = 0; tid < num_threads; tid++) { + for (ThreadID tid = 0; tid < num_threads; tid++) { unsigned tid_TCStatus = tc->readRegOtherThread(MipsISA::TCStatus + Ctrl_Base_DepTag, tid); unsigned tid_TCHalt = tc->readRegOtherThread(MipsISA::TCHalt + Ctrl_Base_DepTag, diff --git a/src/arch/mips/regfile.cc b/src/arch/mips/regfile.cc index 663115bb6..908302866 100644 --- a/src/arch/mips/regfile.cc +++ b/src/arch/mips/regfile.cc @@ -54,7 +54,9 @@ void RegFile::clear() miscRegFile.clear(); } -void RegFile::reset(std::string core_name, unsigned num_threads, unsigned num_vpes) +void +RegFile::reset(std::string core_name, ThreadID num_threads, + unsigned num_vpes) { bzero(&intRegFile, sizeof(intRegFile)); bzero(&floatRegFile, sizeof(floatRegFile)); @@ -71,24 +73,27 @@ Fault RegFile::setIntReg(int intReg, const IntReg &val) return intRegFile.setReg(intReg, val); } -MiscReg RegFile::readMiscRegNoEffect(int miscReg, unsigned tid = 0) +MiscReg +RegFile::readMiscRegNoEffect(int miscReg, ThreadID tid = 0) { return miscRegFile.readRegNoEffect(miscReg, tid); } -MiscReg RegFile::readMiscReg(int miscReg, ThreadContext *tc, - unsigned tid = 0) +MiscReg +RegFile::readMiscReg(int miscReg, ThreadContext *tc, ThreadID tid = 0) { return miscRegFile.readReg(miscReg, tc, tid); } -void RegFile::setMiscRegNoEffect(int miscReg, const MiscReg &val, unsigned tid = 0) +void +RegFile::setMiscRegNoEffect(int miscReg, const MiscReg &val, ThreadID tid = 0) { miscRegFile.setRegNoEffect(miscReg, val, tid); } -void RegFile::setMiscReg(int miscReg, const MiscReg &val, - ThreadContext * tc, unsigned tid = 0) +void +RegFile::setMiscReg(int miscReg, const MiscReg &val, + ThreadContext *tc, ThreadID tid = 0) { miscRegFile.setReg(miscReg, val, tc, tid); } diff --git a/src/arch/mips/regfile/float_regfile.cc b/src/arch/mips/regfile/float_regfile.cc index 122d7c229..2b32bd3af 100644 --- a/src/arch/mips/regfile/float_regfile.cc +++ b/src/arch/mips/regfile/float_regfile.cc @@ -42,7 +42,7 @@ FloatRegFile::clear() } double -FloatRegFile::readReg(int floatReg, int width, unsigned tid) +FloatRegFile::readReg(int floatReg, int width, ThreadID tid) { switch(width) { @@ -65,7 +65,7 @@ FloatRegFile::readReg(int floatReg, int width, unsigned tid) } FloatRegBits -FloatRegFile::readRegBits(int floatReg, int width, unsigned tid) +FloatRegFile::readRegBits(int floatReg, int width, ThreadID tid) { if (floatReg < NumFloatArchRegs - 1) { switch(width) @@ -88,9 +88,9 @@ FloatRegFile::readRegBits(int floatReg, int width, unsigned tid) } Fault -FloatRegFile::setReg(int floatReg, const FloatRegVal &val, int width, unsigned tid) +FloatRegFile::setReg(int floatReg, const FloatRegVal &val, int width, + ThreadID tid) { - using namespace std; switch(width) { case SingleWidth: @@ -118,10 +118,9 @@ FloatRegFile::setReg(int floatReg, const FloatRegVal &val, int width, unsigned t } Fault -FloatRegFile::setRegBits(int floatReg, const FloatRegBits &val, int width, unsigned tid) +FloatRegFile::setRegBits(int floatReg, const FloatRegBits &val, int width, + ThreadID tid) { - using namespace std; - switch(width) { case SingleWidth: diff --git a/src/arch/mips/regfile/float_regfile.hh b/src/arch/mips/regfile/float_regfile.hh index 7a95d0568..afe6701c5 100644 --- a/src/arch/mips/regfile/float_regfile.hh +++ b/src/arch/mips/regfile/float_regfile.hh @@ -88,10 +88,12 @@ namespace MipsISA static const int regWidth = SingleWidth; void clear(); - double readReg(int floatReg, int width, unsigned tid = 0); - FloatRegBits readRegBits(int floatReg, int width, unsigned tid = 0); - Fault setReg(int floatReg, const FloatRegVal &val, int width, unsigned tid = 0); - Fault setRegBits(int floatReg, const FloatRegBits &val, int width, unsigned tid = 0); + double readReg(int floatReg, int width, ThreadID tid = 0); + FloatRegBits readRegBits(int floatReg, int width, ThreadID tid = 0); + Fault setReg(int floatReg, const FloatRegVal &val, int width, + ThreadID tid = 0); + Fault setRegBits(int floatReg, const FloatRegBits &val, int width, + ThreadID tid = 0); void serialize(std::ostream &os); void unserialize(Checkpoint *cp, const std::string §ion); diff --git a/src/arch/mips/regfile/misc_regfile.cc b/src/arch/mips/regfile/misc_regfile.cc index a00bf166e..aee4fab4d 100644 --- a/src/arch/mips/regfile/misc_regfile.cc +++ b/src/arch/mips/regfile/misc_regfile.cc @@ -131,7 +131,7 @@ MiscRegFile::clear(unsigned tid_or_vpn) } void -MiscRegFile::expandForMultithreading(unsigned num_threads, unsigned num_vpes) +MiscRegFile::expandForMultithreading(ThreadID num_threads, unsigned num_vpes) { // Initialize all Per-VPE regs uint32_t per_vpe_regs[] = { VPEControl, VPEConf0, VPEConf1, YQMask, @@ -180,7 +180,7 @@ int MiscRegFile:: getDataAsid() } //@TODO: Use MIPS STYLE CONSTANTS (e.g. TCHALT_H instead of TCH_H) void -MiscRegFile::reset(std::string core_name, unsigned num_threads, +MiscRegFile::reset(std::string core_name, ThreadID num_threads, unsigned num_vpes, BaseCPU *_cpu) { DPRINTF(MipsPRA, "Resetting CP0 State with %i TCs and %i VPEs\n", @@ -378,7 +378,7 @@ MiscRegFile::reset(std::string core_name, unsigned num_threads, setRegNoEffect(VPEConf0, vpe_conf0); // TCBind - for (int tid = 0; tid < num_threads; tid++) { + for (ThreadID tid = 0; tid < num_threads; tid++) { MiscReg tc_bind = readRegNoEffect(TCBind, tid); replaceBits(tc_bind, TCB_CUR_TC_HI, TCB_CUR_TC_LO, tid); setRegNoEffect(TCBind, tc_bind, tid); @@ -387,7 +387,7 @@ MiscRegFile::reset(std::string core_name, unsigned num_threads, MiscReg tc_halt = readRegNoEffect(TCHalt); replaceBits(tc_halt, TCH_H, 0); setRegNoEffect(TCHalt, tc_halt); - /*for (int tid = 1; tid < num_threads; tid++) { + /*for (ThreadID tid = 1; tid < num_threads; tid++) { // Set TCHalt Halt bit to 1 for all other threads tc_halt = readRegNoEffect(TCHalt, tid); replaceBits(tc_halt, TCH_H, 1); @@ -401,7 +401,7 @@ MiscRegFile::reset(std::string core_name, unsigned num_threads, setRegNoEffect(TCStatus, tc_status); // Set Dynamically Allocatable bit to 1 for all other threads - for (int tid = 1; tid < num_threads; tid++) { + for (ThreadID tid = 1; tid < num_threads; tid++) { tc_status = readRegNoEffect(TCStatus, tid); replaceBits(tc_status, TCSTATUS_DA, 1); setRegNoEffect(TCStatus, tc_status, tid); @@ -439,14 +439,14 @@ MiscRegFile::reset(std::string core_name, unsigned num_threads, } inline unsigned -MiscRegFile::getVPENum(unsigned tid) +MiscRegFile::getVPENum(ThreadID tid) { unsigned tc_bind = miscRegFile[TCBind - Ctrl_Base_DepTag][tid]; return bits(tc_bind, TCB_CUR_VPE_HI, TCB_CUR_VPE_LO); } MiscReg -MiscRegFile::readRegNoEffect(int reg_idx, unsigned tid) +MiscRegFile::readRegNoEffect(int reg_idx, ThreadID tid) { int misc_reg = reg_idx - Ctrl_Base_DepTag; unsigned reg_sel = (bankType[misc_reg] == perThreadContext) @@ -461,8 +461,7 @@ MiscRegFile::readRegNoEffect(int reg_idx, unsigned tid) // Status to TCStatus depending on current thread //template MiscReg -MiscRegFile::readReg(int reg_idx, - ThreadContext *tc, unsigned tid) +MiscRegFile::readReg(int reg_idx, ThreadContext *tc, ThreadID tid) { int misc_reg = reg_idx - Ctrl_Base_DepTag; unsigned reg_sel = (bankType[misc_reg] == perThreadContext) @@ -481,7 +480,7 @@ MiscRegFile::readReg(int reg_idx, } void -MiscRegFile::setRegNoEffect(int reg_idx, const MiscReg &val, unsigned tid) +MiscRegFile::setRegNoEffect(int reg_idx, const MiscReg &val, ThreadID tid) { int misc_reg = reg_idx - Ctrl_Base_DepTag; unsigned reg_sel = (bankType[misc_reg] == perThreadContext) @@ -494,7 +493,7 @@ MiscRegFile::setRegNoEffect(int reg_idx, const MiscReg &val, unsigned tid) miscRegFile[misc_reg][reg_sel] = val; } void -MiscRegFile::setRegMask(int reg_idx, const MiscReg &val, unsigned tid) +MiscRegFile::setRegMask(int reg_idx, const MiscReg &val, ThreadID tid) { // return; int misc_reg = reg_idx - Ctrl_Base_DepTag; @@ -513,7 +512,7 @@ MiscRegFile::setRegMask(int reg_idx, const MiscReg &val, unsigned tid) //template void MiscRegFile::setReg(int reg_idx, const MiscReg &val, - ThreadContext *tc, unsigned tid) + ThreadContext *tc, ThreadID tid) { int misc_reg = reg_idx - Ctrl_Base_DepTag; int reg_sel = (bankType[misc_reg] == perThreadContext) @@ -575,9 +574,9 @@ MiscRegFile::updateCPU() // /////////////////////////////////////////////////////////////////// unsigned mvp_conf0 = readRegNoEffect(MVPConf0); - unsigned num_threads = bits(mvp_conf0, MVPC0_PTC_HI, MVPC0_PTC_LO) + 1; + ThreadID num_threads = bits(mvp_conf0, MVPC0_PTC_HI, MVPC0_PTC_LO) + 1; - for (int tid = 0; tid < num_threads; tid++) { + for (ThreadID tid = 0; tid < num_threads; tid++) { MiscReg tc_status = readRegNoEffect(TCStatus, tid); MiscReg tc_halt = readRegNoEffect(TCHalt, tid); diff --git a/src/arch/mips/regfile/misc_regfile.hh b/src/arch/mips/regfile/misc_regfile.hh index 0daf8f718..633ea7efb 100644 --- a/src/arch/mips/regfile/misc_regfile.hh +++ b/src/arch/mips/regfile/misc_regfile.hh @@ -75,12 +75,12 @@ namespace MipsISA void clear(unsigned tid_or_vpn = 0); - void reset(std::string core_name, unsigned num_threads, + void reset(std::string core_name, ThreadID num_threads, unsigned num_vpes, BaseCPU *_cpu); - void expandForMultithreading(unsigned num_threads, unsigned num_vpes); + void expandForMultithreading(ThreadID num_threads, unsigned num_vpes); - inline unsigned getVPENum(unsigned tid); + unsigned getVPENum(ThreadID tid); ////////////////////////////////////////////////////////// // @@ -90,21 +90,21 @@ namespace MipsISA ////////////////////////////////////////////////////////// //@TODO: MIPS MT's register view automatically connects // Status to TCStatus depending on current thread - void updateCP0ReadView(int misc_reg, unsigned tid) { } - MiscReg readRegNoEffect(int misc_reg, unsigned tid = 0); + void updateCP0ReadView(int misc_reg, ThreadID tid) { } + MiscReg readRegNoEffect(int misc_reg, ThreadID tid = 0); //template MiscReg readReg(int misc_reg, - ThreadContext *tc, unsigned tid = 0); + ThreadContext *tc, ThreadID tid = 0); MiscReg filterCP0Write(int misc_reg, int reg_sel, const MiscReg &val); - void setRegMask(int misc_reg, const MiscReg &val, unsigned tid = 0); + void setRegMask(int misc_reg, const MiscReg &val, ThreadID tid = 0); void setRegNoEffect(int misc_reg, const MiscReg &val, - unsigned tid = 0); + ThreadID tid = 0); //template void setReg(int misc_reg, const MiscReg &val, - ThreadContext *tc, unsigned tid = 0); + ThreadContext *tc, ThreadID tid = 0); int getInstAsid(); int getDataAsid(); diff --git a/src/arch/mips/regfile/regfile.cc b/src/arch/mips/regfile/regfile.cc index a1c8eab6a..975fad963 100644 --- a/src/arch/mips/regfile/regfile.cc +++ b/src/arch/mips/regfile/regfile.cc @@ -46,7 +46,8 @@ RegFile::clear() } void -RegFile::reset(std::string core_name, unsigned num_threads, unsigned num_vpes, BaseCPU *_cpu) +RegFile::reset(std::string core_name, ThreadID num_threads, unsigned num_vpes, + BaseCPU *_cpu) { bzero(&intRegFile, sizeof(intRegFile)); bzero(&floatRegFile, sizeof(floatRegFile)); @@ -66,27 +67,26 @@ RegFile::setIntReg(int intReg, const IntReg &val) } MiscReg -RegFile::readMiscRegNoEffect(int miscReg, unsigned tid) +RegFile::readMiscRegNoEffect(int miscReg, ThreadID tid) { return miscRegFile.readRegNoEffect(miscReg, tid); } MiscReg -RegFile::readMiscReg(int miscReg, ThreadContext *tc, - unsigned tid) +RegFile::readMiscReg(int miscReg, ThreadContext *tc, ThreadID tid) { return miscRegFile.readReg(miscReg, tc, tid); } void -RegFile::setMiscRegNoEffect(int miscReg, const MiscReg &val, unsigned tid) +RegFile::setMiscRegNoEffect(int miscReg, const MiscReg &val, ThreadID tid) { miscRegFile.setRegNoEffect(miscReg, val, tid); } void RegFile::setMiscReg(int miscReg, const MiscReg &val, - ThreadContext * tc, unsigned tid) + ThreadContext *tc, ThreadID tid) { miscRegFile.setReg(miscReg, val, tc, tid); } diff --git a/src/arch/mips/regfile/regfile.hh b/src/arch/mips/regfile/regfile.hh index ebf793396..91951b078 100644 --- a/src/arch/mips/regfile/regfile.hh +++ b/src/arch/mips/regfile/regfile.hh @@ -61,19 +61,21 @@ namespace MipsISA public: void clear(); - void reset(std::string core_name, unsigned num_threads, unsigned num_vpes, BaseCPU *_cpu); + void reset(std::string core_name, ThreadID num_threads, + unsigned num_vpes, BaseCPU *_cpu); MiscRegFile *getMiscRegFilePtr(); IntReg readIntReg(int intReg); Fault setIntReg(int intReg, const IntReg &val); - MiscReg readMiscRegNoEffect(int miscReg, unsigned tid = 0); + MiscReg readMiscRegNoEffect(int miscReg, ThreadID tid = 0); MiscReg readMiscReg(int miscReg, ThreadContext *tc, - unsigned tid = 0); - void setMiscRegNoEffect(int miscReg, const MiscReg &val, unsigned tid = 0); + ThreadID tid = 0); + void setMiscRegNoEffect(int miscReg, const MiscReg &val, + ThreadID tid = 0); void setMiscReg(int miscReg, const MiscReg &val, - ThreadContext * tc, unsigned tid = 0); + ThreadContext *tc, ThreadID tid = 0); FloatRegVal readFloatReg(int floatReg); FloatRegVal readFloatReg(int floatReg, int width); diff --git a/src/base/types.hh b/src/base/types.hh index 6d8d3095c..1a6db9fbb 100644 --- a/src/base/types.hh +++ b/src/base/types.hh @@ -68,4 +68,10 @@ typedef uint64_t Addr; const Addr MaxAddr = (Addr)-1; +/** + * Thread index/ID type + */ +typedef int16_t ThreadID; +const ThreadID InvalidThreadID = (ThreadID)-1; + #endif // __BASE_TYPES_HH__ diff --git a/src/cpu/base.cc b/src/cpu/base.cc index 78ff4f998..e4cb79344 100644 --- a/src/cpu/base.cc +++ b/src/cpu/base.cc @@ -100,12 +100,12 @@ CPUProgressEvent::description() const BaseCPU::BaseCPU(Params *p) : MemObject(p), clock(p->clock), instCnt(0), _cpuId(p->cpu_id), interrupts(p->interrupts), - number_of_threads(p->numThreads), system(p->system), + numThreads(p->numThreads), system(p->system), phase(p->phase) #else BaseCPU::BaseCPU(Params *p) : MemObject(p), clock(p->clock), _cpuId(p->cpu_id), - number_of_threads(p->numThreads), system(p->system), + numThreads(p->numThreads), system(p->system), phase(p->phase) #endif { @@ -121,22 +121,23 @@ BaseCPU::BaseCPU(Params *p) DPRINTF(SyscallVerbose, "Constructing CPU with id %d\n", _cpuId); - if (number_of_threads > maxThreadsPerCPU) - maxThreadsPerCPU = number_of_threads; + if (numThreads > maxThreadsPerCPU) + maxThreadsPerCPU = numThreads; // allocate per-thread instruction-based event queues - comInstEventQueue = new EventQueue *[number_of_threads]; - for (int i = 0; i < number_of_threads; ++i) - comInstEventQueue[i] = new EventQueue("instruction-based event queue"); + comInstEventQueue = new EventQueue *[numThreads]; + for (ThreadID tid = 0; tid < numThreads; ++tid) + comInstEventQueue[tid] = + new EventQueue("instruction-based event queue"); // // set up instruction-count-based termination events, if any // if (p->max_insts_any_thread != 0) { const char *cause = "a thread reached the max instruction count"; - for (int i = 0; i < number_of_threads; ++i) { + for (ThreadID tid = 0; tid < numThreads; ++tid) { Event *event = new SimLoopExitEvent(cause, 0); - comInstEventQueue[i]->schedule(event, p->max_insts_any_thread); + comInstEventQueue[tid]->schedule(event, p->max_insts_any_thread); } } @@ -147,26 +148,26 @@ BaseCPU::BaseCPU(Params *p) // decrement this when triggered; simulation will terminate // when counter reaches 0 int *counter = new int; - *counter = number_of_threads; - for (int i = 0; i < number_of_threads; ++i) { + *counter = numThreads; + for (ThreadID tid = 0; tid < numThreads; ++tid) { Event *event = new CountedExitEvent(cause, *counter); - comInstEventQueue[i]->schedule(event, p->max_insts_any_thread); + comInstEventQueue[tid]->schedule(event, p->max_insts_any_thread); } } // allocate per-thread load-based event queues - comLoadEventQueue = new EventQueue *[number_of_threads]; - for (int i = 0; i < number_of_threads; ++i) - comLoadEventQueue[i] = new EventQueue("load-based event queue"); + comLoadEventQueue = new EventQueue *[numThreads]; + for (ThreadID tid = 0; tid < numThreads; ++tid) + comLoadEventQueue[tid] = new EventQueue("load-based event queue"); // // set up instruction-count-based termination events, if any // if (p->max_loads_any_thread != 0) { const char *cause = "a thread reached the max load count"; - for (int i = 0; i < number_of_threads; ++i) { + for (ThreadID tid = 0; tid < numThreads; ++tid) { Event *event = new SimLoopExitEvent(cause, 0); - comLoadEventQueue[i]->schedule(event, p->max_loads_any_thread); + comLoadEventQueue[tid]->schedule(event, p->max_loads_any_thread); } } @@ -176,10 +177,10 @@ BaseCPU::BaseCPU(Params *p) // decrement this when triggered; simulation will terminate // when counter reaches 0 int *counter = new int; - *counter = number_of_threads; - for (int i = 0; i < number_of_threads; ++i) { + *counter = numThreads; + for (ThreadID tid = 0; tid < numThreads; ++tid) { Event *event = new CountedExitEvent(cause, *counter); - comLoadEventQueue[i]->schedule(event, p->max_loads_all_threads); + comLoadEventQueue[tid]->schedule(event, p->max_loads_all_threads); } } @@ -289,8 +290,9 @@ BaseCPU::nextCycle(Tick begin_tick) void BaseCPU::registerThreadContexts() { - for (int i = 0; i < threadContexts.size(); ++i) { - ThreadContext *tc = threadContexts[i]; + ThreadID size = threadContexts.size(); + for (ThreadID tid = 0; tid < size; ++tid) { + ThreadContext *tc = threadContexts[tid]; /** This is so that contextId and cpuId match where there is a * 1cpu:1context relationship. Otherwise, the order of registration @@ -299,7 +301,7 @@ BaseCPU::registerThreadContexts() * cpu 0 has the lowest thread contexts and cpu N has the highest, but * I'll just do this for now */ - if (number_of_threads == 1) + if (numThreads == 1) tc->setContextId(system->registerThreadContext(tc, _cpuId)); else tc->setContextId(system->registerThreadContext(tc)); @@ -313,9 +315,10 @@ BaseCPU::registerThreadContexts() int BaseCPU::findContext(ThreadContext *tc) { - for (int i = 0; i < threadContexts.size(); ++i) { - if (tc == threadContexts[i]) - return i; + ThreadID size = threadContexts.size(); + for (ThreadID tid = 0; tid < size; ++tid) { + if (tc == threadContexts[tid]) + return tid; } return 0; } @@ -337,7 +340,8 @@ BaseCPU::takeOverFrom(BaseCPU *oldCPU, Port *ic, Port *dc) _cpuId = oldCPU->cpuId(); - for (int i = 0; i < threadContexts.size(); ++i) { + ThreadID size = threadContexts.size(); + for (ThreadID i = 0; i < size; ++i) { ThreadContext *newTC = threadContexts[i]; ThreadContext *oldTC = oldCPU->threadContexts[i]; @@ -361,7 +365,7 @@ BaseCPU::takeOverFrom(BaseCPU *oldCPU, Port *ic, Port *dc) interrupts = oldCPU->interrupts; interrupts->setCPU(this); - for (int i = 0; i < threadContexts.size(); ++i) + for (ThreadID i = 0; i < size; ++i) threadContexts[i]->profileClear(); if (profileEvent) @@ -393,7 +397,8 @@ BaseCPU::ProfileEvent::ProfileEvent(BaseCPU *_cpu, Tick _interval) void BaseCPU::ProfileEvent::process() { - for (int i = 0, size = cpu->threadContexts.size(); i < size; ++i) { + ThreadID size = cpu->threadContexts.size(); + for (ThreadID i = 0; i < size; ++i) { ThreadContext *tc = cpu->threadContexts[i]; tc->profileSample(); } diff --git a/src/cpu/base.hh b/src/cpu/base.hh index e63960cc1..441d9b5dd 100644 --- a/src/cpu/base.hh +++ b/src/cpu/base.hh @@ -213,7 +213,7 @@ class BaseCPU : public MemObject virtual void startup(); virtual void regStats(); - virtual void activateWhenReady(int tid) {}; + virtual void activateWhenReady(ThreadID tid) {}; void registerThreadContexts(); @@ -229,7 +229,7 @@ class BaseCPU : public MemObject * Number of threads we're actually simulating (<= SMT_MAX_THREADS). * This is a constant for the duration of the simulation. */ - int number_of_threads; + ThreadID numThreads; TheISA::CoreSpecific coreParams; //ISA-Specific Params That Set Up State in Core diff --git a/src/cpu/base_dyn_inst.hh b/src/cpu/base_dyn_inst.hh index ed0054402..c6e57b612 100644 --- a/src/cpu/base_dyn_inst.hh +++ b/src/cpu/base_dyn_inst.hh @@ -168,7 +168,7 @@ class BaseDynInst : public FastAlloc, public RefCounted std::bitset status; /** The thread this instruction is from. */ - short threadNumber; + ThreadID threadNumber; /** data address space ID, for loads & stores. */ short asid; @@ -778,7 +778,7 @@ class BaseDynInst : public FastAlloc, public RefCounted void setASID(short addr_space_id) { asid = addr_space_id; } /** Sets the thread id. */ - void setTid(unsigned tid) { threadNumber = tid; } + void setTid(ThreadID tid) { threadNumber = tid; } /** Sets the pointer to the thread state. */ void setThreadState(ImplState *state) { thread = state; } diff --git a/src/cpu/btb.cc b/src/cpu/btb.cc index 93d6ee768..4c1d441be 100644 --- a/src/cpu/btb.cc +++ b/src/cpu/btb.cc @@ -82,7 +82,7 @@ DefaultBTB::getTag(const Addr &inst_PC) } bool -DefaultBTB::valid(const Addr &inst_PC, unsigned tid) +DefaultBTB::valid(const Addr &inst_PC, ThreadID tid) { unsigned btb_idx = getIndex(inst_PC); @@ -103,7 +103,7 @@ DefaultBTB::valid(const Addr &inst_PC, unsigned tid) // 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, unsigned tid) +DefaultBTB::lookup(const Addr &inst_PC, ThreadID tid) { unsigned btb_idx = getIndex(inst_PC); @@ -121,7 +121,7 @@ DefaultBTB::lookup(const Addr &inst_PC, unsigned tid) } void -DefaultBTB::update(const Addr &inst_PC, const Addr &target, unsigned tid) +DefaultBTB::update(const Addr &inst_PC, const Addr &target, ThreadID tid) { unsigned btb_idx = getIndex(inst_PC); diff --git a/src/cpu/inorder/cpu.cc b/src/cpu/inorder/cpu.cc index e901b475e..24483bf25 100644 --- a/src/cpu/inorder/cpu.cc +++ b/src/cpu/inorder/cpu.cc @@ -29,9 +29,10 @@ * */ -#include "config/full_system.hh" +#include #include "arch/utility.hh" +#include "config/full_system.hh" #include "cpu/exetrace.hh" #include "cpu/activity.hh" #include "cpu/simple_thread.hh" @@ -49,7 +50,6 @@ #include "mem/translating_port.hh" #include "sim/process.hh" #include "sim/stat_control.hh" -#include using namespace std; using namespace TheISA; @@ -74,7 +74,7 @@ InOrderCPU::TickEvent::description() } InOrderCPU::CPUEvent::CPUEvent(InOrderCPU *_cpu, CPUEventType e_type, - Fault fault, unsigned _tid, unsigned _vpe) + Fault fault, ThreadID _tid, unsigned _vpe) : Event(CPU_Tick_Pri), cpu(_cpu) { setEvent(e_type, fault, _tid, _vpe); @@ -175,7 +175,6 @@ InOrderCPU::InOrderCPU(Params *params) switchCount(0), deferRegistration(false/*params->deferRegistration*/), stageTracing(params->stageTracing), - numThreads(params->numThreads), numVirtProcs(1) { cpu_params = params; @@ -204,28 +203,28 @@ InOrderCPU::InOrderCPU(Params *params) fatal("Unable to find port for data.\n"); } - for (int i = 0; i < numThreads; ++i) { - if (i < params->workload.size()) { + for (ThreadID tid = 0; tid < numThreads; ++tid) { + if (tid < params->workload.size()) { DPRINTF(InOrderCPU, "Workload[%i] process is %#x\n", - i, this->thread[i]); - this->thread[i] = new Thread(this, i, params->workload[i], - i); + tid, this->thread[tid]); + this->thread[tid] = + new Thread(this, tid, params->workload[tid], tid); } else { //Allocate Empty thread so M5 can use later //when scheduling threads to CPU Process* dummy_proc = params->workload[0]; - this->thread[i] = new Thread(this, i, dummy_proc, i); + this->thread[tid] = new Thread(this, tid, dummy_proc, tid); } // Setup the TC that will serve as the interface to the threads/CPU. InOrderThreadContext *tc = new InOrderThreadContext; tc->cpu = this; - tc->thread = this->thread[i]; + tc->thread = thread[tid]; // Give the thread the TC. - thread[i]->tc = tc; - thread[i]->setFuncExeInst(0); - globalSeqNum[i] = 1; + thread[tid]->tc = tc; + thread[tid]->setFuncExeInst(0); + globalSeqNum[tid] = 1; // Add the TC to the CPU's list of TC's. this->threadContexts.push_back(tc); @@ -257,7 +256,7 @@ InOrderCPU::InOrderCPU(Params *params) } // Initialize thread specific variables - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { archRegDepMap[tid].setCPU(this); nonSpecInstActive[tid] = false; @@ -434,19 +433,19 @@ InOrderCPU::init() // Set inSyscall so that the CPU doesn't squash when initially // setting up registers. - for (int i = 0; i < number_of_threads; ++i) - thread[i]->inSyscall = true; + for (ThreadID tid = 0; tid < numThreads; ++tid) + thread[tid]->inSyscall = true; #if FULL_SYSTEM - for (int tid=0; tid < number_of_threads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { ThreadContext *src_tc = threadContexts[tid]; TheISA::initCPU(src_tc, src_tc->contextId()); } #endif // Clear inSyscall. - for (int i = 0; i < number_of_threads; ++i) - thread[i]->inSyscall = false; + for (ThreadID tid = 0; tid < numThreads; ++tid) + thread[tid]->inSyscall = false; // Call Initializiation Routine for Resource Pool resPool->init(); @@ -472,21 +471,21 @@ InOrderCPU::getPort(const std::string &if_name, int idx) } void -InOrderCPU::trap(Fault fault, unsigned tid, int delay) +InOrderCPU::trap(Fault fault, ThreadID tid, int delay) { //@ Squash Pipeline during TRAP scheduleCpuEvent(Trap, fault, tid, 0/*vpe*/, delay); } void -InOrderCPU::trapCPU(Fault fault, unsigned tid) +InOrderCPU::trapCPU(Fault fault, ThreadID tid) { fault->invoke(tcBase(tid)); } void InOrderCPU::scheduleCpuEvent(CPUEventType c_event, Fault fault, - unsigned tid, unsigned vpe, unsigned delay) + ThreadID tid, unsigned vpe, unsigned delay) { CPUEvent *cpu_event = new CPUEvent(this, c_event, fault, tid, vpe); @@ -500,26 +499,27 @@ InOrderCPU::scheduleCpuEvent(CPUEventType c_event, Fault fault, } // Broadcast event to the Resource Pool - DynInstPtr dummy_inst = new InOrderDynInst(this, NULL, getNextEventNum(), tid); + DynInstPtr dummy_inst = + new InOrderDynInst(this, NULL, getNextEventNum(), tid); resPool->scheduleEvent(c_event, dummy_inst, 0, 0, tid); } inline bool -InOrderCPU::isThreadActive(unsigned tid) +InOrderCPU::isThreadActive(ThreadID tid) { - list::iterator isActive = std::find( - activeThreads.begin(), activeThreads.end(), tid); + list::iterator isActive = + std::find(activeThreads.begin(), activeThreads.end(), tid); return (isActive != activeThreads.end()); } void -InOrderCPU::activateThread(unsigned tid) +InOrderCPU::activateThread(ThreadID tid) { if (!isThreadActive(tid)) { - DPRINTF(InOrderCPU, "Adding Thread %i to active threads list in CPU.\n", - tid); + DPRINTF(InOrderCPU, + "Adding Thread %i to active threads list in CPU.\n", tid); activeThreads.push_back(tid); wakeCPU(); @@ -527,15 +527,15 @@ InOrderCPU::activateThread(unsigned tid) } void -InOrderCPU::deactivateThread(unsigned tid) +InOrderCPU::deactivateThread(ThreadID tid) { DPRINTF(InOrderCPU, "[tid:%i]: Calling deactivate thread.\n", tid); if (isThreadActive(tid)) { DPRINTF(InOrderCPU,"[tid:%i]: Removing from active threads list\n", tid); - list::iterator thread_it = std::find(activeThreads.begin(), - activeThreads.end(), tid); + list::iterator thread_it = + std::find(activeThreads.begin(), activeThreads.end(), tid); removePipelineStalls(*thread_it); @@ -546,7 +546,7 @@ InOrderCPU::deactivateThread(unsigned tid) } void -InOrderCPU::removePipelineStalls(unsigned tid) +InOrderCPU::removePipelineStalls(ThreadID tid) { DPRINTF(InOrderCPU,"[tid:%i]: Removing all pipeline stalls\n", tid); @@ -557,16 +557,16 @@ InOrderCPU::removePipelineStalls(unsigned tid) } bool -InOrderCPU::isThreadInCPU(unsigned tid) +InOrderCPU::isThreadInCPU(ThreadID tid) { - list::iterator isCurrent = std::find( - currentThreads.begin(), currentThreads.end(), tid); + list::iterator isCurrent = + std::find(currentThreads.begin(), currentThreads.end(), tid); return (isCurrent != currentThreads.end()); } void -InOrderCPU::addToCurrentThreads(unsigned tid) +InOrderCPU::addToCurrentThreads(ThreadID tid) { if (!isThreadInCPU(tid)) { DPRINTF(InOrderCPU, "Adding Thread %i to current threads list in CPU.\n", @@ -576,22 +576,22 @@ InOrderCPU::addToCurrentThreads(unsigned tid) } void -InOrderCPU::removeFromCurrentThreads(unsigned tid) +InOrderCPU::removeFromCurrentThreads(ThreadID tid) { if (isThreadInCPU(tid)) { - DPRINTF(InOrderCPU, "Adding Thread %i to current threads list in CPU.\n", - tid); - list::iterator isCurrent = std::find( - currentThreads.begin(), currentThreads.end(), tid); + DPRINTF(InOrderCPU, + "Adding Thread %i to current threads list in CPU.\n", tid); + list::iterator isCurrent = + std::find(currentThreads.begin(), currentThreads.end(), tid); currentThreads.erase(isCurrent); } } bool -InOrderCPU::isThreadSuspended(unsigned tid) +InOrderCPU::isThreadSuspended(ThreadID tid) { - list::iterator isSuspended = std::find( - suspendedThreads.begin(), suspendedThreads.end(), tid); + list::iterator isSuspended = + std::find(suspendedThreads.begin(), suspendedThreads.end(), tid); return (isSuspended!= suspendedThreads.end()); } @@ -612,7 +612,7 @@ InOrderCPU::enableVPEs(unsigned vpe) DPRINTF(InOrderCPU, "[vpe:%i]: Enabling Concurrent Execution " "virtual processors %i", vpe); - list::iterator thread_it = currentThreads.begin(); + list::iterator thread_it = currentThreads.begin(); while (thread_it != currentThreads.end()) { if (!isThreadSuspended(*thread_it)) { @@ -623,7 +623,7 @@ InOrderCPU::enableVPEs(unsigned vpe) } void -InOrderCPU::disableVirtProcElement(unsigned tid, unsigned vpe) +InOrderCPU::disableVirtProcElement(ThreadID tid, unsigned vpe) { DPRINTF(InOrderCPU, "[vpe:%i]: Scheduling " "Disabling of concurrent virtual processor execution", @@ -633,16 +633,16 @@ InOrderCPU::disableVirtProcElement(unsigned tid, unsigned vpe) } void -InOrderCPU::disableVPEs(unsigned tid, unsigned vpe) +InOrderCPU::disableVPEs(ThreadID tid, unsigned vpe) { DPRINTF(InOrderCPU, "[vpe:%i]: Disabling Concurrent Execution of " "virtual processors %i", vpe); unsigned base_vpe = TheISA::getVirtProcNum(tcBase(tid)); - list::iterator thread_it = activeThreads.begin(); + list::iterator thread_it = activeThreads.begin(); - std::vector::iterator> removeList; + vector::iterator> removeList; while (thread_it != activeThreads.end()) { if (base_vpe != vpe) { @@ -672,7 +672,7 @@ InOrderCPU::enableThreads(unsigned vpe) DPRINTF(InOrderCPU, "[vpe:%i]: Enabling Multithreading on " "virtual processor %i", vpe); - list::iterator thread_it = currentThreads.begin(); + list::iterator thread_it = currentThreads.begin(); while (thread_it != currentThreads.end()) { if (TheISA::getVirtProcNum(tcBase(*thread_it)) == vpe) { @@ -684,7 +684,7 @@ InOrderCPU::enableThreads(unsigned vpe) } } void -InOrderCPU::disableMultiThreading(unsigned tid, unsigned vpe) +InOrderCPU::disableMultiThreading(ThreadID tid, unsigned vpe) { // Schedule event to take place at end of cycle DPRINTF(InOrderCPU, "[tid:%i]: Scheduling Disable Multithreading on " @@ -694,14 +694,14 @@ InOrderCPU::disableMultiThreading(unsigned tid, unsigned vpe) } void -InOrderCPU::disableThreads(unsigned tid, unsigned vpe) +InOrderCPU::disableThreads(ThreadID tid, unsigned vpe) { DPRINTF(InOrderCPU, "[tid:%i]: Disabling Multithreading on " "virtual processor %i", tid, vpe); - list::iterator thread_it = activeThreads.begin(); + list::iterator thread_it = activeThreads.begin(); - std::vector::iterator> removeList; + vector::iterator> removeList; while (thread_it != activeThreads.end()) { if (TheISA::getVirtProcNum(tcBase(*thread_it)) == vpe) { @@ -722,8 +722,8 @@ InOrderCPU::updateThreadPriority() { //DEFAULT TO ROUND ROBIN SCHEME //e.g. Move highest priority to end of thread list - list::iterator list_begin = activeThreads.begin(); - list::iterator list_end = activeThreads.end(); + list::iterator list_begin = activeThreads.begin(); + list::iterator list_end = activeThreads.end(); unsigned high_thread = *list_begin; @@ -737,7 +737,7 @@ inline void InOrderCPU::tickThreadStats() { /** Keep track of cycles that each thread is active */ - list::iterator thread_it = activeThreads.begin(); + list::iterator thread_it = activeThreads.begin(); while (thread_it != activeThreads.end()) { threadCycles[*thread_it]++; thread_it++; @@ -750,7 +750,7 @@ InOrderCPU::tickThreadStats() } void -InOrderCPU::activateContext(unsigned tid, int delay) +InOrderCPU::activateContext(ThreadID tid, int delay) { DPRINTF(InOrderCPU,"[tid:%i]: Activating ...\n", tid); @@ -765,27 +765,27 @@ InOrderCPU::activateContext(unsigned tid, int delay) void -InOrderCPU::suspendContext(unsigned tid, int delay) +InOrderCPU::suspendContext(ThreadID tid, int delay) { scheduleCpuEvent(SuspendThread, NoFault, tid, 0/*vpe*/, delay); //_status = Idle; } void -InOrderCPU::suspendThread(unsigned tid) +InOrderCPU::suspendThread(ThreadID tid) { DPRINTF(InOrderCPU,"[tid: %i]: Suspended ...\n", tid); deactivateThread(tid); } void -InOrderCPU::deallocateContext(unsigned tid, int delay) +InOrderCPU::deallocateContext(ThreadID tid, int delay) { scheduleCpuEvent(DeallocateThread, NoFault, tid, 0/*vpe*/, delay); } void -InOrderCPU::deallocateThread(unsigned tid) +InOrderCPU::deallocateThread(ThreadID tid) { DPRINTF(InOrderCPU,"[tid:%i]: Deallocating ...", tid); @@ -797,7 +797,7 @@ InOrderCPU::deallocateThread(unsigned tid) } void -InOrderCPU::squashThreadInPipeline(unsigned tid) +InOrderCPU::squashThreadInPipeline(ThreadID tid) { //Squash all instructions in each stage for (int stNum=NumStages - 1; stNum >= 0 ; stNum--) { @@ -806,7 +806,7 @@ InOrderCPU::squashThreadInPipeline(unsigned tid) } void -InOrderCPU::haltContext(unsigned tid, int delay) +InOrderCPU::haltContext(ThreadID tid, int delay) { DPRINTF(InOrderCPU, "[tid:%i]: Halt context called.\n", tid); @@ -817,13 +817,13 @@ InOrderCPU::haltContext(unsigned tid, int delay) } void -InOrderCPU::insertThread(unsigned tid) +InOrderCPU::insertThread(ThreadID tid) { panic("Unimplemented Function\n."); } void -InOrderCPU::removeThread(unsigned tid) +InOrderCPU::removeThread(ThreadID tid) { DPRINTF(InOrderCPU, "Removing Thread %i from CPU.\n", tid); @@ -838,97 +838,98 @@ InOrderCPU::getPipeStage(int stage_num) void -InOrderCPU::activateWhenReady(int tid) +InOrderCPU::activateWhenReady(ThreadID tid) { panic("Unimplemented Function\n."); } uint64_t -InOrderCPU::readPC(unsigned tid) +InOrderCPU::readPC(ThreadID tid) { return PC[tid]; } void -InOrderCPU::setPC(Addr new_PC, unsigned tid) +InOrderCPU::setPC(Addr new_PC, ThreadID tid) { PC[tid] = new_PC; } uint64_t -InOrderCPU::readNextPC(unsigned tid) +InOrderCPU::readNextPC(ThreadID tid) { return nextPC[tid]; } void -InOrderCPU::setNextPC(uint64_t new_NPC, unsigned tid) +InOrderCPU::setNextPC(uint64_t new_NPC, ThreadID tid) { nextPC[tid] = new_NPC; } uint64_t -InOrderCPU::readNextNPC(unsigned tid) +InOrderCPU::readNextNPC(ThreadID tid) { return nextNPC[tid]; } void -InOrderCPU::setNextNPC(uint64_t new_NNPC, unsigned tid) +InOrderCPU::setNextNPC(uint64_t new_NNPC, ThreadID tid) { nextNPC[tid] = new_NNPC; } uint64_t -InOrderCPU::readIntReg(int reg_idx, unsigned tid) +InOrderCPU::readIntReg(int reg_idx, ThreadID tid) { return intRegFile[tid].readReg(reg_idx); } FloatReg -InOrderCPU::readFloatReg(int reg_idx, unsigned tid, int width) +InOrderCPU::readFloatReg(int reg_idx, ThreadID tid, int width) { return floatRegFile[tid].readReg(reg_idx, width); } FloatRegBits -InOrderCPU::readFloatRegBits(int reg_idx, unsigned tid, int width) +InOrderCPU::readFloatRegBits(int reg_idx, ThreadID tid, int width) {; return floatRegFile[tid].readRegBits(reg_idx, width); } void -InOrderCPU::setIntReg(int reg_idx, uint64_t val, unsigned tid) +InOrderCPU::setIntReg(int reg_idx, uint64_t val, ThreadID tid) { intRegFile[tid].setReg(reg_idx, val); } void -InOrderCPU::setFloatReg(int reg_idx, FloatReg val, unsigned tid, int width) +InOrderCPU::setFloatReg(int reg_idx, FloatReg val, ThreadID tid, int width) { floatRegFile[tid].setReg(reg_idx, val, width); } void -InOrderCPU::setFloatRegBits(int reg_idx, FloatRegBits val, unsigned tid, int width) +InOrderCPU::setFloatRegBits(int reg_idx, FloatRegBits val, ThreadID tid, + int width) { floatRegFile[tid].setRegBits(reg_idx, val, width); } uint64_t -InOrderCPU::readRegOtherThread(unsigned reg_idx, unsigned tid) +InOrderCPU::readRegOtherThread(unsigned reg_idx, ThreadID tid) { // If Default value is set, then retrieve target thread - if (tid == -1) { + if (tid == InvalidThreadID) { tid = TheISA::getTargetThread(tcBase(tid)); } @@ -943,10 +944,11 @@ InOrderCPU::readRegOtherThread(unsigned reg_idx, unsigned tid) } } void -InOrderCPU::setRegOtherThread(unsigned reg_idx, const MiscReg &val, unsigned tid) +InOrderCPU::setRegOtherThread(unsigned reg_idx, const MiscReg &val, + ThreadID tid) { // If Default value is set, then retrieve target thread - if (tid == -1) { + if (tid == InvalidThreadID) { tid = TheISA::getTargetThread(tcBase(tid)); } @@ -962,25 +964,25 @@ InOrderCPU::setRegOtherThread(unsigned reg_idx, const MiscReg &val, unsigned tid } MiscReg -InOrderCPU::readMiscRegNoEffect(int misc_reg, unsigned tid) +InOrderCPU::readMiscRegNoEffect(int misc_reg, ThreadID tid) { return miscRegFile.readRegNoEffect(misc_reg, tid); } MiscReg -InOrderCPU::readMiscReg(int misc_reg, unsigned tid) +InOrderCPU::readMiscReg(int misc_reg, ThreadID tid) { return miscRegFile.readReg(misc_reg, tcBase(tid), tid); } void -InOrderCPU::setMiscRegNoEffect(int misc_reg, const MiscReg &val, unsigned tid) +InOrderCPU::setMiscRegNoEffect(int misc_reg, const MiscReg &val, ThreadID tid) { miscRegFile.setRegNoEffect(misc_reg, val, tid); } void -InOrderCPU::setMiscReg(int misc_reg, const MiscReg &val, unsigned tid) +InOrderCPU::setMiscReg(int misc_reg, const MiscReg &val, ThreadID tid) { miscRegFile.setReg(misc_reg, val, tcBase(tid), tid); } @@ -989,7 +991,7 @@ InOrderCPU::setMiscReg(int misc_reg, const MiscReg &val, unsigned tid) InOrderCPU::ListIt InOrderCPU::addInst(DynInstPtr &inst) { - int tid = inst->readTid(); + ThreadID tid = inst->readTid(); instList[tid].push_back(inst); @@ -997,7 +999,7 @@ InOrderCPU::addInst(DynInstPtr &inst) } void -InOrderCPU::instDone(DynInstPtr inst, unsigned tid) +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 a fork or after an exception @@ -1069,8 +1071,7 @@ InOrderCPU::removeInst(DynInstPtr &inst) } void -InOrderCPU::removeInstsUntil(const InstSeqNum &seq_num, - unsigned tid) +InOrderCPU::removeInstsUntil(const InstSeqNum &seq_num, ThreadID tid) { //assert(!instList[tid].empty()); @@ -1099,7 +1100,7 @@ InOrderCPU::removeInstsUntil(const InstSeqNum &seq_num, inline void -InOrderCPU::squashInstIt(const ListIt &instIt, const unsigned &tid) +InOrderCPU::squashInstIt(const ListIt &instIt, ThreadID tid) { if ((*instIt)->threadNumber == tid) { DPRINTF(InOrderCPU, "Squashing instruction, " @@ -1126,7 +1127,7 @@ InOrderCPU::cleanUpRemovedInsts() (*removeList.front())->readPC()); DynInstPtr inst = *removeList.front(); - int tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; // Make Sure Resource Schedule Is Emptied Out ThePipeline::ResSchedule *inst_sched = &inst->resSched; @@ -1234,7 +1235,7 @@ InOrderCPU::wakeCPU() } void -InOrderCPU::syscall(int64_t callnum, int tid) +InOrderCPU::syscall(int64_t callnum, ThreadID tid) { DPRINTF(InOrderCPU, "[tid:%i] Executing syscall().\n\n", tid); diff --git a/src/cpu/inorder/cpu.hh b/src/cpu/inorder/cpu.hh index 55b04f6a9..faf37382f 100644 --- a/src/cpu/inorder/cpu.hh +++ b/src/cpu/inorder/cpu.hh @@ -41,6 +41,7 @@ #include "arch/isa_traits.hh" #include "base/statistics.hh" #include "base/timebuf.hh" +#include "base/types.hh" #include "config/full_system.hh" #include "cpu/activity.hh" #include "cpu/base.hh" @@ -185,17 +186,18 @@ class InOrderCPU : public BaseCPU public: CPUEventType cpuEventType; - unsigned tid; + ThreadID tid; unsigned vpe; Fault fault; public: /** Constructs a CPU event. */ CPUEvent(InOrderCPU *_cpu, CPUEventType e_type, Fault fault, - unsigned _tid, unsigned _vpe); + ThreadID _tid, unsigned _vpe); /** Set Type of Event To Be Scheduled */ - void setEvent(CPUEventType e_type, Fault _fault, unsigned _tid, unsigned _vpe) + void setEvent(CPUEventType e_type, Fault _fault, ThreadID _tid, + unsigned _vpe) { fault = _fault; cpuEventType = e_type; @@ -217,7 +219,7 @@ class InOrderCPU : public BaseCPU }; /** Schedule a CPU Event */ - void scheduleCpuEvent(CPUEventType cpu_event, Fault fault, unsigned tid, + void scheduleCpuEvent(CPUEventType cpu_event, Fault fault, ThreadID tid, unsigned vpe, unsigned delay = 0); public: @@ -293,29 +295,29 @@ class InOrderCPU : public BaseCPU /** trap() - sets up a trap event on the cpuTraps to handle given fault. * trapCPU() - Traps to handle given fault */ - void trap(Fault fault, unsigned tid, int delay = 0); - void trapCPU(Fault fault, unsigned tid); + void trap(Fault fault, ThreadID tid, int delay = 0); + void trapCPU(Fault fault, ThreadID tid); /** Setup CPU to insert a thread's context */ - void insertThread(unsigned tid); + void insertThread(ThreadID tid); /** Remove all of a thread's context from CPU */ - void removeThread(unsigned tid); + void removeThread(ThreadID tid); /** Add Thread to Active Threads List. */ - void activateContext(unsigned tid, int delay = 0); - void activateThread(unsigned tid); + void activateContext(ThreadID tid, int delay = 0); + void activateThread(ThreadID tid); /** Remove Thread from Active Threads List */ - void suspendContext(unsigned tid, int delay = 0); - void suspendThread(unsigned tid); + void suspendContext(ThreadID tid, int delay = 0); + void suspendThread(ThreadID tid); /** Remove Thread from Active Threads List && * Remove Thread Context from CPU. */ - void deallocateContext(unsigned tid, int delay = 0); - void deallocateThread(unsigned tid); - void deactivateThread(unsigned tid); + void deallocateContext(ThreadID tid, int delay = 0); + void deallocateThread(ThreadID tid); + void deactivateThread(ThreadID tid); PipelineStage* getPipeStage(int stage_num); @@ -329,30 +331,30 @@ class InOrderCPU : public BaseCPU /** Remove Thread from Active Threads List && * Remove Thread Context from CPU. */ - void haltContext(unsigned tid, int delay = 0); + void haltContext(ThreadID tid, int delay = 0); - void removePipelineStalls(unsigned tid); + void removePipelineStalls(ThreadID tid); - void squashThreadInPipeline(unsigned tid); + void squashThreadInPipeline(ThreadID tid); /// Notify the CPU to enable a virtual processor element. virtual void enableVirtProcElement(unsigned vpe); void enableVPEs(unsigned vpe); /// Notify the CPU to disable a virtual processor element. - virtual void disableVirtProcElement(unsigned tid, unsigned vpe); - void disableVPEs(unsigned tid, unsigned vpe); + virtual void disableVirtProcElement(ThreadID tid, unsigned vpe); + void disableVPEs(ThreadID tid, unsigned vpe); /// Notify the CPU that multithreading is enabled. virtual void enableMultiThreading(unsigned vpe); void enableThreads(unsigned vpe); /// Notify the CPU that multithreading is disabled. - virtual void disableMultiThreading(unsigned tid, unsigned vpe); - void disableThreads(unsigned tid, unsigned vpe); + virtual void disableMultiThreading(ThreadID tid, unsigned vpe); + void disableThreads(ThreadID tid, unsigned vpe); /** Activate a Thread When CPU Resources are Available. */ - void activateWhenReady(int tid); + void activateWhenReady(ThreadID tid); /** Add or Remove a Thread Context in the CPU. */ void doContextSwitch(); @@ -365,19 +367,19 @@ class InOrderCPU : public BaseCPU { /*pipelineStage[stage_idx]->switchToActive();*/ } /** Get the current instruction sequence number, and increment it. */ - InstSeqNum getAndIncrementInstSeq(unsigned tid) + InstSeqNum getAndIncrementInstSeq(ThreadID tid) { return globalSeqNum[tid]++; } /** Get the current instruction sequence number, and increment it. */ - InstSeqNum nextInstSeqNum(unsigned tid) + InstSeqNum nextInstSeqNum(ThreadID tid) { return globalSeqNum[tid]; } /** Increment Instruction Sequence Number */ - void incrInstSeqNum(unsigned tid) + void incrInstSeqNum(ThreadID tid) { globalSeqNum[tid]++; } /** Set Instruction Sequence Number */ - void setInstSeqNum(unsigned tid, InstSeqNum seq_num) + void setInstSeqNum(ThreadID tid, InstSeqNum seq_num) { globalSeqNum[tid] = seq_num; } @@ -389,73 +391,76 @@ class InOrderCPU : public BaseCPU } /** Get instruction asid. */ - int getInstAsid(unsigned tid) + int getInstAsid(ThreadID tid) { return thread[tid]->getInstAsid(); } /** Get data asid. */ - int getDataAsid(unsigned tid) + int getDataAsid(ThreadID tid) { return thread[tid]->getDataAsid(); } /** Register file accessors */ - uint64_t readIntReg(int reg_idx, unsigned tid); + uint64_t readIntReg(int reg_idx, ThreadID tid); - FloatReg readFloatReg(int reg_idx, unsigned tid, + FloatReg readFloatReg(int reg_idx, ThreadID tid, int width = TheISA::SingleWidth); - FloatRegBits readFloatRegBits(int reg_idx, unsigned tid, + FloatRegBits readFloatRegBits(int reg_idx, ThreadID tid, int width = TheISA::SingleWidth); - void setIntReg(int reg_idx, uint64_t val, unsigned tid); + void setIntReg(int reg_idx, uint64_t val, ThreadID tid); - void setFloatReg(int reg_idx, FloatReg val, unsigned tid, + void setFloatReg(int reg_idx, FloatReg val, ThreadID tid, int width = TheISA::SingleWidth); - void setFloatRegBits(int reg_idx, FloatRegBits val, unsigned tid, + void setFloatRegBits(int reg_idx, FloatRegBits val, ThreadID tid, int width = TheISA::SingleWidth); /** Reads a miscellaneous register. */ - MiscReg readMiscRegNoEffect(int misc_reg, unsigned tid = 0); + MiscReg readMiscRegNoEffect(int misc_reg, ThreadID tid = 0); /** Reads a misc. register, including any side effects the read * might have as defined by the architecture. */ - MiscReg readMiscReg(int misc_reg, unsigned tid = 0); + MiscReg readMiscReg(int misc_reg, ThreadID tid = 0); /** Sets a miscellaneous register. */ - void setMiscRegNoEffect(int misc_reg, const MiscReg &val, unsigned tid = 0); + void setMiscRegNoEffect(int misc_reg, const MiscReg &val, + ThreadID tid = 0); /** Sets a misc. register, including any side effects the write * might have as defined by the architecture. */ - void setMiscReg(int misc_reg, const MiscReg &val, unsigned tid = 0); + void setMiscReg(int misc_reg, const MiscReg &val, ThreadID tid = 0); /** Reads a int/fp/misc reg. from another thread depending on ISA-defined * target thread */ - uint64_t readRegOtherThread(unsigned misc_reg, unsigned tid = -1); + uint64_t readRegOtherThread(unsigned misc_reg, + ThreadID tid = InvalidThreadID); /** Sets a int/fp/misc reg. from another thread depending on an ISA-defined * target thread */ - void setRegOtherThread(unsigned misc_reg, const MiscReg &val, unsigned tid); + void setRegOtherThread(unsigned misc_reg, const MiscReg &val, + ThreadID tid); /** Reads the commit PC of a specific thread. */ - uint64_t readPC(unsigned tid); + uint64_t readPC(ThreadID tid); /** Sets the commit PC of a specific thread. */ - void setPC(Addr new_PC, unsigned tid); + void setPC(Addr new_PC, ThreadID tid); /** Reads the next PC of a specific thread. */ - uint64_t readNextPC(unsigned tid); + uint64_t readNextPC(ThreadID tid); /** Sets the next PC of a specific thread. */ - void setNextPC(uint64_t val, unsigned tid); + void setNextPC(uint64_t val, ThreadID tid); /** Reads the next NPC of a specific thread. */ - uint64_t readNextNPC(unsigned tid); + uint64_t readNextNPC(ThreadID tid); /** Sets the next NPC of a specific thread. */ - void setNextNPC(uint64_t val, unsigned tid); + void setNextNPC(uint64_t val, ThreadID tid); /** Function to add instruction onto the head of the list of the * instructions. Used when new instructions are fetched. @@ -463,7 +468,7 @@ class InOrderCPU : public BaseCPU ListIt addInst(DynInstPtr &inst); /** Function to tell the CPU that an instruction has completed. */ - void instDone(DynInstPtr inst, unsigned tid); + void instDone(DynInstPtr inst, ThreadID tid); /** Add Instructions to the CPU Remove List*/ void addToRemoveList(DynInstPtr &inst); @@ -472,10 +477,10 @@ class InOrderCPU : public BaseCPU void removeInst(DynInstPtr &inst); /** Remove all instructions younger than the given sequence number. */ - void removeInstsUntil(const InstSeqNum &seq_num,unsigned tid); + void removeInstsUntil(const InstSeqNum &seq_num,ThreadID tid); /** Removes the instruction pointed to by the iterator. */ - inline void squashInstIt(const ListIt &instIt, const unsigned &tid); + inline void squashInstIt(const ListIt &instIt, ThreadID tid); /** Cleans up all instructions on the instruction remove list. */ void cleanUpRemovedInsts(); @@ -513,7 +518,7 @@ class InOrderCPU : public BaseCPU void writeHint(DynInstPtr inst); /** Executes a syscall.*/ - void syscall(int64_t callnum, int tid); + void syscall(int64_t callnum, ThreadID tid); public: /** Per-Thread List of all the instructions in flight. */ @@ -553,24 +558,24 @@ class InOrderCPU : public BaseCPU /** Last Cycle that the CPU squashed instruction end. */ Tick lastSquashCycle[ThePipeline::MaxThreads]; - std::list fetchPriorityList; + std::list fetchPriorityList; protected: /** Active Threads List */ - std::list activeThreads; + std::list activeThreads; /** Current Threads List */ - std::list currentThreads; + std::list currentThreads; /** Suspended Threads List */ - std::list suspendedThreads; + std::list suspendedThreads; /** Thread Status Functions (Unused Currently) */ - bool isThreadInCPU(unsigned tid); - bool isThreadActive(unsigned tid); - bool isThreadSuspended(unsigned tid); - void addToCurrentThreads(unsigned tid); - void removeFromCurrentThreads(unsigned tid); + bool isThreadInCPU(ThreadID tid); + bool isThreadActive(ThreadID tid); + bool isThreadSuspended(ThreadID tid); + void addToCurrentThreads(ThreadID tid); + void removeFromCurrentThreads(ThreadID tid); private: /** The activity recorder; used to tell if the CPU has any @@ -583,7 +588,7 @@ class InOrderCPU : public BaseCPU void readFunctional(Addr addr, uint32_t &buffer); /** Number of Active Threads in the CPU */ - int numActiveThreads() { return activeThreads.size(); } + ThreadID numActiveThreads() { return activeThreads.size(); } /** Records that there was time buffer activity this cycle. */ void activityThisCycle() { activityRec.activity(); } @@ -600,7 +605,7 @@ class InOrderCPU : public BaseCPU void wakeCPU(); /** Gets a free thread id. Use if thread ids change across system. */ - int getFreeTid(); + ThreadID getFreeTid(); // LL/SC debug functionality unsigned stCondFails; @@ -608,7 +613,7 @@ class InOrderCPU : public BaseCPU unsigned setStCondFailures(unsigned st_fails) { return stCondFails = st_fails; } /** Returns a pointer to a thread context. */ - ThreadContext *tcBase(unsigned tid = 0) + ThreadContext *tcBase(ThreadID tid = 0) { return thread[tid]->getTC(); } @@ -618,8 +623,8 @@ class InOrderCPU : public BaseCPU { Counter total(0); - for (int i=0; i < thread.size(); i++) - total += thread[i]->numInst; + for (ThreadID tid = 0; tid < thread.size(); tid++) + total += thread[tid]->numInst; return total; } @@ -657,9 +662,6 @@ class InOrderCPU : public BaseCPU /** The cycle that the CPU was last running, used for statistics. */ Tick lastRunningCycle; - /** Number of Threads the CPU can process */ - unsigned numThreads; - /** Number of Virtual Processors the CPU can process */ unsigned numVirtProcs; diff --git a/src/cpu/inorder/first_stage.cc b/src/cpu/inorder/first_stage.cc index cc0af13f9..8bd703c56 100644 --- a/src/cpu/inorder/first_stage.cc +++ b/src/cpu/inorder/first_stage.cc @@ -43,7 +43,7 @@ FirstStage::FirstStage(Params *params, unsigned stage_num) : PipelineStage(params, stage_num), numFetchingThreads(1), fetchPolicy(FirstStage::RoundRobin) { - for(int tid=0; tid < this->numThreads; tid++) { + for(ThreadID tid = 0; tid < this->numThreads; tid++) { stageStatus[tid] = Running; } } @@ -60,7 +60,7 @@ FirstStage::setCPU(InOrderCPU *cpu_ptr) void -FirstStage::squash(InstSeqNum squash_seq_num, unsigned tid) +FirstStage::squash(InstSeqNum squash_seq_num, ThreadID tid) { // Set status to squashing. //stageStatus[tid] = Squashing; @@ -96,17 +96,17 @@ FirstStage::squash(InstSeqNum squash_seq_num, unsigned tid) void FirstStage::processStage(bool &status_change) { - list::iterator threads = (*activeThreads).begin(); + list::iterator threads = activeThreads->begin(); //Check stall and squash signals. - while (threads != (*activeThreads).end()) { - unsigned tid = *threads++; + while (threads != activeThreads->end()) { + ThreadID tid = *threads++; status_change = checkSignalsAndUpdate(tid) || status_change; } for (int threadFetched = 0; threadFetched < numFetchingThreads; threadFetched++) { - int tid = getFetchingThread(fetchPolicy); + ThreadID tid = getFetchingThread(fetchPolicy); if (tid >= 0) { DPRINTF(InOrderStage, "Processing [tid:%i]\n",tid); @@ -120,7 +120,7 @@ FirstStage::processStage(bool &status_change) //@TODO: Note in documentation, that when you make a pipeline stage change, then //make sure you change the first stage too void -FirstStage::processInsts(unsigned tid) +FirstStage::processInsts(ThreadID tid) { bool all_reqs_completed = true; @@ -194,7 +194,7 @@ FirstStage::processInsts(unsigned tid) } } -int +ThreadID FirstStage::getFetchingThread(FetchPriority &fetch_priority) { if (numThreads > 1) { @@ -207,28 +207,28 @@ FirstStage::getFetchingThread(FetchPriority &fetch_priority) return roundRobin(); default: - return -1; + return InvalidThreadID; } } else { - int tid = *((*activeThreads).begin()); + ThreadID tid = *activeThreads->begin(); if (stageStatus[tid] == Running || stageStatus[tid] == Idle) { return tid; } else { - return -1; + return InvalidThreadID; } } } -int +ThreadID FirstStage::roundRobin() { - list::iterator pri_iter = (*fetchPriorityList).begin(); - list::iterator end = (*fetchPriorityList).end(); + list::iterator pri_iter = fetchPriorityList->begin(); + list::iterator end = fetchPriorityList->end(); - int high_pri; + ThreadID high_pri; while (pri_iter != end) { high_pri = *pri_iter; @@ -238,8 +238,8 @@ FirstStage::roundRobin() if (stageStatus[high_pri] == Running || stageStatus[high_pri] == Idle) { - (*fetchPriorityList).erase(pri_iter); - (*fetchPriorityList).push_back(high_pri); + fetchPriorityList->erase(pri_iter); + fetchPriorityList->push_back(high_pri); return high_pri; } @@ -247,5 +247,5 @@ FirstStage::roundRobin() pri_iter++; } - return -1; + return InvalidThreadID; } diff --git a/src/cpu/inorder/first_stage.hh b/src/cpu/inorder/first_stage.hh index 55914c85c..2a69678e4 100644 --- a/src/cpu/inorder/first_stage.hh +++ b/src/cpu/inorder/first_stage.hh @@ -56,10 +56,10 @@ class FirstStage : public PipelineStage { void processStage(bool &status_change); /** Process All Instructions Available */ - void processInsts(unsigned tid); + void processInsts(ThreadID tid); /** Squash Instructions Above a Seq. Num */ - void squash(InstSeqNum squash_seq_num, unsigned tid); + void squash(InstSeqNum squash_seq_num, ThreadID tid); /** There are no insts. coming from previous stages, so there is * no need to sort insts here @@ -69,7 +69,7 @@ class FirstStage : public PipelineStage { /** There are no skidBuffers for the first stage. So * just use an empty function. */ - void skidInsert(unsigned tid) { } + void skidInsert(ThreadID tid) { } /** The number of fetching threads in the CPU */ int numFetchingThreads; @@ -85,13 +85,13 @@ class FirstStage : public PipelineStage { FetchPriority fetchPolicy; /** List that has the threads organized by priority. */ - std::list *fetchPriorityList; + std::list *fetchPriorityList; /** Return the next fetching thread */ - int getFetchingThread(FetchPriority &fetch_priority); + ThreadID getFetchingThread(FetchPriority &fetch_priority); - /** Return next thred given Round Robin Policy for Thread Fetching */ - int roundRobin(); + /** Return next thread given Round Robin Policy for Thread Fetching */ + ThreadID roundRobin(); }; #endif // __CPU_INORDER_FIRST_STAGE_HH__ diff --git a/src/cpu/inorder/inorder_cpu_builder.cc b/src/cpu/inorder/inorder_cpu_builder.cc index 0088a3bd9..5ee7b31db 100644 --- a/src/cpu/inorder/inorder_cpu_builder.cc +++ b/src/cpu/inorder/inorder_cpu_builder.cc @@ -42,7 +42,7 @@ InOrderCPU * InOrderCPUParams::create() { - int actual_num_threads = + ThreadID actual_num_threads = (numThreads >= workload.size()) ? numThreads : workload.size(); if (workload.size() == 0) { diff --git a/src/cpu/inorder/inorder_dyn_inst.cc b/src/cpu/inorder/inorder_dyn_inst.cc index ed63d9148..ee2e5500e 100644 --- a/src/cpu/inorder/inorder_dyn_inst.cc +++ b/src/cpu/inorder/inorder_dyn_inst.cc @@ -66,7 +66,7 @@ InOrderDynInst::InOrderDynInst(TheISA::ExtMachInst machInst, Addr inst_PC, InOrderDynInst::InOrderDynInst(InOrderCPU *cpu, InOrderThreadState *state, InstSeqNum seq_num, - unsigned tid, + ThreadID tid, unsigned _asid) : traceData(NULL), cpu(cpu) { @@ -385,7 +385,7 @@ InOrderDynInst::setFloatRegBitsSrc(int idx, uint64_t val) /** Reads a integer register. */ IntReg -InOrderDynInst::readIntRegOperand(const StaticInst *si, int idx, unsigned tid) +InOrderDynInst::readIntRegOperand(const StaticInst *si, int idx, ThreadID tid) { DPRINTF(InOrderDynInst, "[tid:%i]: [sn:%i] Source Value %i read as %#x.\n", threadNumber, seqNum, idx, instSrc[idx].integer); @@ -479,7 +479,7 @@ InOrderDynInst::setMiscRegOperand(const StaticInst *si, int idx, } MiscReg -InOrderDynInst::readRegOtherThread(unsigned reg_idx, int tid) +InOrderDynInst::readRegOtherThread(unsigned reg_idx, ThreadID tid) { if (tid == -1) { tid = TheISA::getTargetThread(this->cpu->tcBase(threadNumber)); @@ -553,9 +553,10 @@ InOrderDynInst::setMiscReg(int misc_reg, const MiscReg &val) } void -InOrderDynInst::setRegOtherThread(unsigned reg_idx, const MiscReg &val, int tid) +InOrderDynInst::setRegOtherThread(unsigned reg_idx, const MiscReg &val, + ThreadID tid) { - if (tid == -1) { + if (tid == InvalidThreadID) { tid = TheISA::getTargetThread(this->cpu->tcBase(threadNumber)); } diff --git a/src/cpu/inorder/inorder_dyn_inst.hh b/src/cpu/inorder/inorder_dyn_inst.hh index 52465a712..031d882ee 100644 --- a/src/cpu/inorder/inorder_dyn_inst.hh +++ b/src/cpu/inorder/inorder_dyn_inst.hh @@ -112,8 +112,8 @@ class InOrderDynInst : public FastAlloc, public RefCounted * @param cpu Pointer to the instruction's CPU. * NOTE: Must set Binary Instrution through Member Function */ - InOrderDynInst(InOrderCPU *cpu, InOrderThreadState *state, InstSeqNum seq_num, - unsigned tid, unsigned asid = 0); + InOrderDynInst(InOrderCPU *cpu, InOrderThreadState *state, + InstSeqNum seq_num, ThreadID tid, unsigned asid = 0); /** BaseDynInst constructor given a StaticInst pointer. * @param _staticInst The StaticInst for this BaseDynInst. @@ -347,7 +347,7 @@ class InOrderDynInst : public FastAlloc, public RefCounted short readTid() { return threadNumber; } /** Sets the thread id. */ - void setTid(unsigned tid) { threadNumber = tid; } + void setTid(ThreadID tid) { threadNumber = tid; } void setVpn(int id) { virtProcNumber = id; } @@ -829,7 +829,7 @@ class InOrderDynInst : public FastAlloc, public RefCounted * language (which is why the name isnt readIntSrc(...)) Note: That * the source reg. value is set using the setSrcReg() function. */ - IntReg readIntRegOperand(const StaticInst *si, int idx, unsigned tid=0); + IntReg readIntRegOperand(const StaticInst *si, int idx, ThreadID tid = 0); FloatReg readFloatRegOperand(const StaticInst *si, int idx, int width = TheISA::SingleWidth); TheISA::FloatRegBits readFloatRegOperandBits(const StaticInst *si, int idx, @@ -881,8 +881,10 @@ class InOrderDynInst : public FastAlloc, public RefCounted void setMiscRegOperand(const StaticInst *si, int idx, const MiscReg &val); void setMiscRegOperandNoEffect(const StaticInst *si, int idx, const MiscReg &val); - virtual uint64_t readRegOtherThread(unsigned idx, int tid = -1); - virtual void setRegOtherThread(unsigned idx, const uint64_t &val, int tid = -1); + virtual uint64_t readRegOtherThread(unsigned idx, + ThreadID tid = InvalidThreadID); + virtual void setRegOtherThread(unsigned idx, const uint64_t &val, + ThreadID tid = InvalidThreadID); /** Sets the number of consecutive store conditional failures. */ void setStCondFailures(unsigned sc_failures) diff --git a/src/cpu/inorder/pipeline_stage.cc b/src/cpu/inorder/pipeline_stage.cc index dd31716eb..46b1cbad0 100644 --- a/src/cpu/inorder/pipeline_stage.cc +++ b/src/cpu/inorder/pipeline_stage.cc @@ -49,7 +49,7 @@ PipelineStage::PipelineStage(Params *params, unsigned stage_num) void PipelineStage::init(Params *params) { - for(int tid=0; tid < numThreads; tid++) { + for(ThreadID tid = 0; tid < numThreads; tid++) { stageStatus[tid] = Idle; for (int stNum = 0; stNum < NumStages; stNum++) { @@ -162,7 +162,7 @@ PipelineStage::setNextStageQueue(TimeBuffer *next_stage_ptr) void -PipelineStage::setActiveThreads(list *at_ptr) +PipelineStage::setActiveThreads(list *at_ptr) { DPRINTF(InOrderStage, "Setting active threads list pointer.\n"); activeThreads = at_ptr; @@ -194,20 +194,20 @@ PipelineStage::takeOverFrom() _status = Inactive; // Be sure to reset state and clear out any old instructions. - for (int i = 0; i < numThreads; ++i) { - stageStatus[i] = Idle; + for (ThreadID tid = 0; tid < numThreads; ++tid) { + stageStatus[tid] = Idle; for (int stNum = 0; stNum < NumStages; stNum++) { - stalls[i].stage[stNum] = false; + stalls[tid].stage[stNum] = false; } - stalls[i].resources.clear(); + stalls[tid].resources.clear(); - while (!insts[i].empty()) - insts[i].pop(); + while (!insts[tid].empty()) + insts[tid].pop(); - while (!skidBuffer[i].empty()) - skidBuffer[i].pop(); + while (!skidBuffer[tid].empty()) + skidBuffer[tid].pop(); } wroteToTimeBuffer = false; } @@ -215,7 +215,7 @@ PipelineStage::takeOverFrom() bool -PipelineStage::checkStall(unsigned tid) const +PipelineStage::checkStall(ThreadID tid) const { bool ret_val = false; @@ -243,7 +243,7 @@ PipelineStage::checkStall(unsigned tid) const void -PipelineStage::removeStalls(unsigned tid) +PipelineStage::removeStalls(ThreadID tid) { for (int stNum = 0; stNum < NumStages; stNum++) { stalls[tid].stage[stNum] = false; @@ -258,13 +258,13 @@ PipelineStage::prevStageInstsValid() } bool -PipelineStage::isBlocked(unsigned tid) +PipelineStage::isBlocked(ThreadID tid) { return stageStatus[tid] == Blocked; } bool -PipelineStage::block(unsigned tid) +PipelineStage::block(ThreadID tid) { DPRINTF(InOrderStage, "[tid:%d]: Blocking, sending block signal back to previous stages.\n", tid); @@ -293,7 +293,7 @@ PipelineStage::block(unsigned tid) } void -PipelineStage::blockDueToBuffer(unsigned tid) +PipelineStage::blockDueToBuffer(ThreadID tid) { DPRINTF(InOrderStage, "[tid:%d]: Blocking instructions from passing to next stage.\n", tid); @@ -308,7 +308,7 @@ PipelineStage::blockDueToBuffer(unsigned tid) } bool -PipelineStage::unblock(unsigned tid) +PipelineStage::unblock(ThreadID tid) { // Stage is done unblocking only if the skid buffer is empty. if (skidBuffer[tid].empty()) { @@ -329,7 +329,7 @@ PipelineStage::unblock(unsigned tid) } void -PipelineStage::squashDueToBranch(DynInstPtr &inst, unsigned tid) +PipelineStage::squashDueToBranch(DynInstPtr &inst, ThreadID tid) { if (cpu->squashSeqNum[tid] < inst->seqNum && cpu->lastSquashCycle[tid] == curTick){ @@ -367,8 +367,7 @@ PipelineStage::squashDueToBranch(DynInstPtr &inst, unsigned tid) } void -PipelineStage::squashPrevStageInsts(InstSeqNum squash_seq_num, - unsigned tid) +PipelineStage::squashPrevStageInsts(InstSeqNum squash_seq_num, ThreadID tid) { DPRINTF(InOrderStage, "[tid:%i]: Removing instructions from " "incoming stage queue.\n", tid); @@ -387,7 +386,7 @@ PipelineStage::squashPrevStageInsts(InstSeqNum squash_seq_num, } void -PipelineStage::squash(InstSeqNum squash_seq_num, unsigned tid) +PipelineStage::squash(InstSeqNum squash_seq_num, ThreadID tid) { // Set status to squashing. stageStatus[tid] = Squashing; @@ -450,7 +449,7 @@ PipelineStage::canSendInstToStage(unsigned stage_num) } void -PipelineStage::skidInsert(unsigned tid) +PipelineStage::skidInsert(ThreadID tid) { DynInstPtr inst = NULL; @@ -484,9 +483,9 @@ PipelineStage::skidSize() bool PipelineStage::skidsEmpty() { - list::iterator threads = (*activeThreads).begin(); + list::iterator threads = activeThreads->begin(); - while (threads != (*activeThreads).end()) { + while (threads != activeThreads->end()) { if (!skidBuffer[*threads++].empty()) return false; } @@ -501,12 +500,10 @@ PipelineStage::updateStatus() { bool any_unblocking = false; - list::iterator threads = (*activeThreads).begin(); + list::iterator threads = activeThreads->begin(); - threads = (*activeThreads).begin(); - - while (threads != (*activeThreads).end()) { - unsigned tid = *threads++; + while (threads != activeThreads->end()) { + ThreadID tid = *threads++; if (stageStatus[tid] == Unblocking) { any_unblocking = true; @@ -561,7 +558,7 @@ PipelineStage::sortInsts() prevStage->insts[i]->readTid(), prevStage->insts[i]->seqNum); - int tid = prevStage->insts[i]->threadNumber; + ThreadID tid = prevStage->insts[i]->threadNumber; DynInstPtr inst = prevStage->insts[i]; @@ -576,7 +573,7 @@ PipelineStage::sortInsts() void -PipelineStage::readStallSignals(unsigned tid) +PipelineStage::readStallSignals(ThreadID tid) { for (int stage_idx = stageNum+1; stage_idx <= lastStallingStage[tid]; stage_idx++) { @@ -597,7 +594,7 @@ PipelineStage::readStallSignals(unsigned tid) bool -PipelineStage::checkSignalsAndUpdate(unsigned tid) +PipelineStage::checkSignalsAndUpdate(ThreadID tid) { // Check if there's a squash signal, squash if there is. // Check stall signals, block if necessary. @@ -691,14 +688,14 @@ PipelineStage::tick() } void -PipelineStage::setResStall(ResReqPtr res_req, unsigned tid) +PipelineStage::setResStall(ResReqPtr res_req, ThreadID tid) { DPRINTF(InOrderStage, "Inserting stall from %s.\n", res_req->res->name()); stalls[tid].resources.push_back(res_req); } void -PipelineStage::unsetResStall(ResReqPtr res_req, unsigned tid) +PipelineStage::unsetResStall(ResReqPtr res_req, ThreadID tid) { // Search through stalls to find stalling request and then // remove it @@ -731,11 +728,11 @@ PipelineStage::unsetResStall(ResReqPtr res_req, unsigned tid) void PipelineStage::processStage(bool &status_change) { - list::iterator threads = (*activeThreads).begin(); + list::iterator threads = activeThreads->begin(); //Check stall and squash signals. - while (threads != (*activeThreads).end()) { - unsigned tid = *threads++; + while (threads != activeThreads->end()) { + ThreadID tid = *threads++; DPRINTF(InOrderStage,"Processing [tid:%i]\n",tid); status_change = checkSignalsAndUpdate(tid) || status_change; @@ -756,7 +753,7 @@ PipelineStage::processStage(bool &status_change) } void -PipelineStage::processThread(bool &status_change, unsigned tid) +PipelineStage::processThread(bool &status_change, ThreadID tid) { // If status is Running or idle, // call stageInsts() @@ -801,7 +798,7 @@ PipelineStage::processThread(bool &status_change, unsigned tid) void -PipelineStage::processInsts(unsigned tid) +PipelineStage::processInsts(ThreadID tid) { // Instructions can come either from the skid buffer or the list of // instructions coming from fetch, depending on stage's status. @@ -888,9 +885,9 @@ bool PipelineStage::processInstSchedule(DynInstPtr inst) { bool last_req_completed = true; - int tid; - - tid = inst->readTid(); +#if TRACING_ON + ThreadID tid = inst->readTid(); +#endif if (inst->nextResStage() == stageNum) { int res_stage_num = inst->nextResStage(); @@ -951,7 +948,7 @@ PipelineStage::sendInstToNextStage(DynInstPtr inst) inst->nextStage++; bool success = false; - int tid = inst->readTid(); + ThreadID tid = inst->readTid(); int next_stage = inst->nextStage; int prev_stage = next_stage - 1; @@ -1012,8 +1009,7 @@ PipelineStage::dumpInsts() { cprintf("Insts in Stage %i skidbuffers\n",stageNum); - for (int tid=0; tid < ThePipeline::MaxThreads; tid++) { - + for (ThreadID tid = 0; tid < ThePipeline::MaxThreads; tid++) { std::queue copy_buff(skidBuffer[tid]); while (!copy_buff.empty()) { diff --git a/src/cpu/inorder/pipeline_stage.hh b/src/cpu/inorder/pipeline_stage.hh index b074639fb..86ee98132 100644 --- a/src/cpu/inorder/pipeline_stage.hh +++ b/src/cpu/inorder/pipeline_stage.hh @@ -80,7 +80,7 @@ class PipelineStage unsigned stageWidth; /** Number of Threads*/ - unsigned numThreads; + ThreadID numThreads; /** Stage status. */ StageStatus _status; @@ -108,7 +108,7 @@ class PipelineStage /** Sets CPU pointer. */ virtual void setCPU(InOrderCPU *cpu_ptr); - virtual void scheduleStageStart(int delay, int tid) { } + virtual void scheduleStageStart(int delay, ThreadID tid) { } /** Sets the main backwards communication time buffer pointer. */ void setTimeBuffer(TimeBuffer *tb_ptr); @@ -120,11 +120,11 @@ class PipelineStage void setNextStageQueue(TimeBuffer *next_stage_ptr); /** Sets pointer to list of active threads. */ - void setActiveThreads(std::list *at_ptr); + void setActiveThreads(std::list *at_ptr); bool nextStageQueueValid(int stage_num); - bool isBlocked(unsigned tid); + bool isBlocked(ThreadID tid); /** Changes the status of this stage to active, and indicates this * to the CPU. @@ -148,13 +148,13 @@ class PipelineStage virtual void tick(); /** Set a resource stall in the pipeline-stage */ - void setResStall(ResReqPtr res_req, unsigned tid); + void setResStall(ResReqPtr res_req, ThreadID tid); /** Unset a resource stall in the pipeline-stage */ - void unsetResStall(ResReqPtr res_req, unsigned tid); + void unsetResStall(ResReqPtr res_req, ThreadID tid); /** Remove all stall signals for a particular thread; */ - virtual void removeStalls(unsigned tid); + virtual void removeStalls(ThreadID tid); /** Is there room in the stage buffer? */ int stageBufferAvail(); @@ -168,14 +168,14 @@ class PipelineStage * change (ie switching from from blocking to unblocking). * @param tid Thread id to stage instructions from. */ - virtual void processThread(bool &status_change, unsigned tid); + virtual void processThread(bool &status_change, ThreadID tid); /** Processes instructions from fetch and passes them on to rename. * Decoding of instructions actually happens when they are created in * fetch, so this function mostly checks if PC-relative branches are * correct. */ - virtual void processInsts(unsigned tid); + virtual void processInsts(ThreadID tid); /** Process all resources on an instruction's resource schedule */ virtual bool processInstSchedule(DynInstPtr inst); @@ -189,7 +189,7 @@ class PipelineStage /** Inserts a thread's instructions into the skid buffer, to be staged * once stage unblocks. */ - virtual void skidInsert(unsigned tid); + virtual void skidInsert(ThreadID tid); /** Total size of all skid buffers */ int skidSize(); @@ -206,13 +206,13 @@ class PipelineStage void sortInsts(); /** Reads all stall signals from the backwards communication timebuffer. */ - virtual void readStallSignals(unsigned tid); + virtual void readStallSignals(ThreadID tid); /** Checks all input signals and updates stage's status appropriately. */ - virtual bool checkSignalsAndUpdate(unsigned tid); + virtual bool checkSignalsAndUpdate(ThreadID tid); /** Checks all stall signals, and returns if any are true. */ - virtual bool checkStall(unsigned tid) const; + virtual bool checkStall(ThreadID tid) const; /** Returns if there any instructions from the previous stage * on this cycle. @@ -223,30 +223,30 @@ class PipelineStage * become blocked. * @return Returns true if there is a status change. */ - virtual bool block(unsigned tid); + virtual bool block(ThreadID tid); - void blockDueToBuffer(unsigned tid); + void blockDueToBuffer(ThreadID tid); /** Switches stage to unblocking if the skid buffer is empty, and * signals back that stage has unblocked. * @return Returns true if there is a status change. */ - virtual bool unblock(unsigned tid); + virtual bool unblock(ThreadID tid); public: /** Squashes if there is a PC-relative branch that was predicted * incorrectly. Sends squash information back to fetch. */ - virtual void squashDueToBranch(DynInstPtr &inst, unsigned tid); + virtual void squashDueToBranch(DynInstPtr &inst, ThreadID tid); /** Squash instructions from stage buffer */ - virtual void squashPrevStageInsts(InstSeqNum squash_seq_num, unsigned tid); + virtual void squashPrevStageInsts(InstSeqNum squash_seq_num, ThreadID tid); /** Squashes due to commit signalling a squash. Changes status to * squashing and clears block/unblock signals as needed. */ - virtual void squash(InstSeqNum squash_num, unsigned tid); + virtual void squash(InstSeqNum squash_num, ThreadID tid); void dumpInsts(); @@ -257,7 +257,7 @@ class PipelineStage Trace::InOrderTrace *tracer; /** List of active thread ids */ - std::list *activeThreads; + std::list *activeThreads; /** Queue of all instructions coming from previous stage on this cycle. */ std::queue insts[ThePipeline::MaxThreads]; diff --git a/src/cpu/inorder/resource.cc b/src/cpu/inorder/resource.cc index 49afefd98..cb5681bc1 100644 --- a/src/cpu/inorder/resource.cc +++ b/src/cpu/inorder/resource.cc @@ -283,7 +283,7 @@ Resource::execute(int slot_idx) } void -Resource::deactivateThread(unsigned tid) +Resource::deactivateThread(ThreadID tid) { // In the most basic case, deactivation means squashing everything // from a particular thread @@ -292,7 +292,8 @@ Resource::deactivateThread(unsigned tid) } void -Resource::squash(DynInstPtr inst, int stage_num, InstSeqNum squash_seq_num, unsigned tid) +Resource::squash(DynInstPtr inst, int stage_num, InstSeqNum squash_seq_num, + ThreadID tid) { std::vector slot_remove_list; diff --git a/src/cpu/inorder/resource.hh b/src/cpu/inorder/resource.hh index 5b4977158..7935e5517 100644 --- a/src/cpu/inorder/resource.hh +++ b/src/cpu/inorder/resource.hh @@ -84,17 +84,17 @@ class Resource { virtual void regStats(); /** Resources that care about thread activation override this. */ - virtual void activateThread(unsigned tid) { } + virtual void activateThread(ThreadID tid) { } /** Deactivate Thread. Default action is to squash all instructions * from deactivated thread. */ - virtual void deactivateThread(unsigned tid); + virtual void deactivateThread(ThreadID tid); /** Resources that care when an instruction has been graduated * can override this */ - virtual void instGraduated(InstSeqNum seq_num,unsigned tid) { } + virtual void instGraduated(InstSeqNum seq_num, ThreadID tid) { } /** Request usage of this resource. Returns a ResourceRequest object * with all the necessary resource information @@ -151,7 +151,8 @@ class Resource { { panic("writeHint undefined for %s", name()); } /** Squash All Requests After This Seq Num */ - virtual void squash(DynInstPtr inst, int stage_num, InstSeqNum squash_seq_num, unsigned tid); + virtual void squash(DynInstPtr inst, int stage_num, + InstSeqNum squash_seq_num, ThreadID tid); /** The number of instructions available that this resource can * can still process @@ -347,8 +348,8 @@ class ResourceRequest int getStageNum() { return stageNum; } /** Set/Get Thread Ids */ - void setTid(unsigned _tid) { tid = _tid; } - int getTid() { return tid; } + void setTid(ThreadID _tid) { tid = _tid; } + ThreadID getTid() { return tid; } /** Instruction this request is for */ DynInstPtr getInst() { return inst; } @@ -393,7 +394,7 @@ class ResourceRequest protected: /** Resource Identification */ - int tid; + ThreadID tid; int stageNum; int resIdx; int slotNum; diff --git a/src/cpu/inorder/resource_pool.9stage.cc b/src/cpu/inorder/resource_pool.9stage.cc index 4a0258e71..05ce91faa 100644 --- a/src/cpu/inorder/resource_pool.9stage.cc +++ b/src/cpu/inorder/resource_pool.9stage.cc @@ -146,7 +146,8 @@ ResourcePool::request(int res_idx, DynInstPtr inst) } void -ResourcePool::squash(DynInstPtr inst, int res_idx, InstSeqNum done_seq_num, int tid) +ResourcePool::squash(DynInstPtr inst, int res_idx, InstSeqNum done_seq_num, + ThreadID tid) { resources[res_idx]->squash(inst, ThePipeline::NumStages-1, done_seq_num, tid); } @@ -165,7 +166,7 @@ ResourcePool::slotsInUse(int res_idx) void ResourcePool::scheduleEvent(InOrderCPU::CPUEventType e_type, DynInstPtr inst, - int delay, int res_idx, int tid) + int delay, int res_idx, ThreadID tid) { assert(delay >= 0); @@ -246,7 +247,8 @@ ResourcePool::unscheduleEvent(int res_idx, DynInstPtr inst) } void -ResourcePool::squashAll(DynInstPtr inst, int stage_num, InstSeqNum done_seq_num, unsigned tid) +ResourcePool::squashAll(DynInstPtr inst, int stage_num, + InstSeqNum done_seq_num, ThreadID tid) { DPRINTF(Resource, "[tid:%i] Stage %i squashing all instructions above [sn:%i].\n", stage_num, tid, done_seq_num); @@ -259,7 +261,7 @@ ResourcePool::squashAll(DynInstPtr inst, int stage_num, InstSeqNum done_seq_num, } void -ResourcePool::activateAll(unsigned tid) +ResourcePool::activateAll(ThreadID tid) { DPRINTF(Resource, "[tid:%i] Broadcasting Thread Activation to all resources.\n", tid); @@ -272,7 +274,7 @@ ResourcePool::activateAll(unsigned tid) } void -ResourcePool::deactivateAll(unsigned tid) +ResourcePool::deactivateAll(ThreadID tid) { DPRINTF(Resource, "[tid:%i] Broadcasting Thread Deactivation to all resources.\n", tid); @@ -285,7 +287,7 @@ ResourcePool::deactivateAll(unsigned tid) } void -ResourcePool::instGraduated(InstSeqNum seq_num,unsigned tid) +ResourcePool::instGraduated(InstSeqNum seq_num, ThreadID tid) { DPRINTF(Resource, "[tid:%i] Broadcasting [sn:%i] graduation to all resources.\n", tid, seq_num); diff --git a/src/cpu/inorder/resource_pool.cc b/src/cpu/inorder/resource_pool.cc index 7bcf2585b..0d78c232b 100644 --- a/src/cpu/inorder/resource_pool.cc +++ b/src/cpu/inorder/resource_pool.cc @@ -164,7 +164,8 @@ ResourcePool::request(int res_idx, DynInstPtr inst) } void -ResourcePool::squash(DynInstPtr inst, int res_idx, InstSeqNum done_seq_num, int tid) +ResourcePool::squash(DynInstPtr inst, int res_idx, InstSeqNum done_seq_num, + ThreadID tid) { resources[res_idx]->squash(inst, ThePipeline::NumStages-1, done_seq_num, tid); } @@ -183,7 +184,7 @@ ResourcePool::slotsInUse(int res_idx) void ResourcePool::scheduleEvent(InOrderCPU::CPUEventType e_type, DynInstPtr inst, - int delay, int res_idx, int tid) + int delay, int res_idx, ThreadID tid) { assert(delay >= 0); @@ -261,7 +262,8 @@ ResourcePool::unscheduleEvent(int res_idx, DynInstPtr inst) } void -ResourcePool::squashAll(DynInstPtr inst, int stage_num, InstSeqNum done_seq_num, unsigned tid) +ResourcePool::squashAll(DynInstPtr inst, int stage_num, + InstSeqNum done_seq_num, ThreadID tid) { DPRINTF(Resource, "[tid:%i] Stage %i squashing all instructions above [sn:%i].\n", stage_num, tid, done_seq_num); @@ -274,7 +276,7 @@ ResourcePool::squashAll(DynInstPtr inst, int stage_num, InstSeqNum done_seq_num, } void -ResourcePool::activateAll(unsigned tid) +ResourcePool::activateAll(ThreadID tid) { DPRINTF(Resource, "[tid:%i] Broadcasting Thread Activation to all resources.\n", tid); @@ -287,7 +289,7 @@ ResourcePool::activateAll(unsigned tid) } void -ResourcePool::deactivateAll(unsigned tid) +ResourcePool::deactivateAll(ThreadID tid) { DPRINTF(Resource, "[tid:%i] Broadcasting Thread Deactivation to all resources.\n", tid); @@ -300,7 +302,7 @@ ResourcePool::deactivateAll(unsigned tid) } void -ResourcePool::instGraduated(InstSeqNum seq_num,unsigned tid) +ResourcePool::instGraduated(InstSeqNum seq_num, ThreadID tid) { DPRINTF(Resource, "[tid:%i] Broadcasting [sn:%i] graduation to all resources.\n", tid, seq_num); @@ -322,7 +324,7 @@ ResourcePool::ResPoolEvent::ResPoolEvent(ResourcePool *_resPool, DynInstPtr _inst, int stage_num, InstSeqNum seq_num, - unsigned _tid) + ThreadID _tid) : Event(CPU_Tick_Pri), resPool(_resPool), eventType(e_type), inst(_inst), seqNum(seq_num), stageNum(stage_num), tid(_tid) diff --git a/src/cpu/inorder/resource_pool.hh b/src/cpu/inorder/resource_pool.hh index 42a07390c..016fae2bf 100644 --- a/src/cpu/inorder/resource_pool.hh +++ b/src/cpu/inorder/resource_pool.hh @@ -81,7 +81,7 @@ class ResourcePool { int stageNum; - unsigned tid; + ThreadID tid; public: /** Constructs a resource event. */ @@ -93,14 +93,14 @@ class ResourcePool { DynInstPtr _inst, int stage_num, InstSeqNum seq_num, - unsigned _tid); + ThreadID _tid); /** Set Type of Event To Be Scheduled */ void setEvent(InOrderCPU::CPUEventType e_type, DynInstPtr _inst, int stage_num, InstSeqNum seq_num, - unsigned _tid) + ThreadID _tid) { eventType = e_type; inst = _inst; @@ -153,20 +153,21 @@ class ResourcePool { ResReqPtr request(int res_idx, DynInstPtr inst); /** Squash The Resource */ - void squash(DynInstPtr inst, int res_idx, InstSeqNum done_seq_num, int tid); + void squash(DynInstPtr inst, int res_idx, InstSeqNum done_seq_num, + ThreadID tid); /** Squash All Resources in Pool after Done Seq. Num */ void squashAll(DynInstPtr inst, int stage_num, - InstSeqNum done_seq_num, unsigned tid); + InstSeqNum done_seq_num, ThreadID tid); /** Activate Thread in all resources */ - void activateAll(unsigned tid); + void activateAll(ThreadID tid); /** De-Activate Thread in all resources */ - void deactivateAll(unsigned tid); + void deactivateAll(ThreadID tid); /** Broadcast graduation to all resources */ - void instGraduated(InstSeqNum seq_num,unsigned tid); + void instGraduated(InstSeqNum seq_num, ThreadID tid); /** The number of instructions available that a resource can * can still process. @@ -178,7 +179,7 @@ class ResourcePool { /** Schedule resource event, regardless of its current state. */ void scheduleEvent(InOrderCPU::CPUEventType e_type, DynInstPtr inst = NULL, - int delay = 0, int res_idx = 0, int tid = 0); + int delay = 0, int res_idx = 0, ThreadID tid = 0); /** UnSchedule resource event, regardless of its current state. */ void unscheduleEvent(int res_idx, DynInstPtr inst); diff --git a/src/cpu/inorder/resources/agen_unit.cc b/src/cpu/inorder/resources/agen_unit.cc index 44cd002ef..44bf8c0ad 100644 --- a/src/cpu/inorder/resources/agen_unit.cc +++ b/src/cpu/inorder/resources/agen_unit.cc @@ -42,10 +42,11 @@ AGENUnit::execute(int slot_num) ResourceRequest* agen_req = reqMap[slot_num]; DynInstPtr inst = reqMap[slot_num]->inst; Fault fault = reqMap[slot_num]->fault; - int tid; +#if TRACING_ON + ThreadID tid = inst->readTid(); +#endif int seq_num = inst->seqNum; - tid = inst->readTid(); agen_req->fault = NoFault; switch (agen_req->cmd) @@ -54,22 +55,27 @@ AGENUnit::execute(int slot_num) { // Load/Store Instruction if (inst->isMemRef()) { - DPRINTF(InOrderAGEN, "[tid:%i] Generating Address for [sn:%i] (%s).\n", - tid, inst->seqNum, inst->staticInst->getName()); + DPRINTF(InOrderAGEN, + "[tid:%i] Generating Address for [sn:%i] (%s).\n", + tid, seq_num, inst->staticInst->getName()); fault = inst->calcEA(); inst->setMemAddr(inst->getEA()); - DPRINTF(InOrderAGEN, "[tid:%i] [sn:%i] Effective address calculated to be: " - "%#x.\n", tid, inst->seqNum, inst->getEA()); + DPRINTF(InOrderAGEN, + "[tid:%i] [sn:%i] Effective address calculated as: %#x\n", + tid, seq_num, inst->getEA()); if (fault == NoFault) { agen_req->done(); } else { - fatal("%s encountered while calculating address for [sn:%i]",fault->name(), seq_num); + fatal("%s encountered while calculating address [sn:%i]", + fault->name(), seq_num); } } else { - DPRINTF(InOrderAGEN, "[tid:] Ignoring non-memory instruction [sn:%i].\n", tid, seq_num); + DPRINTF(InOrderAGEN, + "[tid:] Ignoring non-memory instruction [sn:%i]\n", + tid, seq_num); agen_req->done(); } } diff --git a/src/cpu/inorder/resources/bpred_unit.cc b/src/cpu/inorder/resources/bpred_unit.cc index df6b33792..2ed8586aa 100644 --- a/src/cpu/inorder/resources/bpred_unit.cc +++ b/src/cpu/inorder/resources/bpred_unit.cc @@ -142,7 +142,7 @@ BPredUnit::takeOverFrom() bool -BPredUnit::predict(DynInstPtr &inst, Addr &PC, unsigned tid) +BPredUnit::predict(DynInstPtr &inst, Addr &PC, ThreadID tid) { // See if branch predictor predicts taken. // If so, get its target addr either from the BTB or the RAS. @@ -268,7 +268,7 @@ BPredUnit::predict(DynInstPtr &inst, Addr &PC, unsigned tid) void -BPredUnit::update(const InstSeqNum &done_sn, unsigned tid) +BPredUnit::update(const InstSeqNum &done_sn, ThreadID tid) { DPRINTF(Resource, "BranchPred: [tid:%i]: Commiting branches until sequence" "number %lli.\n", tid, done_sn); @@ -286,7 +286,7 @@ BPredUnit::update(const InstSeqNum &done_sn, unsigned tid) void -BPredUnit::squash(const InstSeqNum &squashed_sn, unsigned tid) +BPredUnit::squash(const InstSeqNum &squashed_sn, ThreadID tid) { History &pred_hist = predHist[tid]; @@ -321,8 +321,8 @@ BPredUnit::squash(const InstSeqNum &squashed_sn, unsigned tid) void BPredUnit::squash(const InstSeqNum &squashed_sn, const Addr &corr_target, - const bool actually_taken, - unsigned tid) + bool actually_taken, + ThreadID tid) { // Now that we know that a branch was mispredicted, we need to undo // all the branches that have been seen up until this branch and diff --git a/src/cpu/inorder/resources/bpred_unit.hh b/src/cpu/inorder/resources/bpred_unit.hh index bd68459d1..abbb70e0d 100644 --- a/src/cpu/inorder/resources/bpred_unit.hh +++ b/src/cpu/inorder/resources/bpred_unit.hh @@ -87,7 +87,7 @@ class BPredUnit * @param tid The thread id. * @return Returns if the branch is taken or not. */ - bool predict(ThePipeline::DynInstPtr &inst, Addr &PC, unsigned tid); + bool predict(ThePipeline::DynInstPtr &inst, Addr &PC, ThreadID tid); // @todo: Rename this function. void BPUncond(void * &bp_history); @@ -98,7 +98,7 @@ class BPredUnit * @param done_sn The sequence number to commit any older updates up until. * @param tid The thread id. */ - void update(const InstSeqNum &done_sn, unsigned tid); + void update(const InstSeqNum &done_sn, ThreadID tid); /** * Squashes all outstanding updates until a given sequence number. @@ -106,7 +106,7 @@ class BPredUnit * until. * @param tid The thread id. */ - void squash(const InstSeqNum &squashed_sn, unsigned tid); + void squash(const InstSeqNum &squashed_sn, ThreadID tid); /** * Squashes all outstanding updates until a given sequence number, and @@ -118,7 +118,7 @@ class BPredUnit * @param tid The thread id. */ void squash(const InstSeqNum &squashed_sn, const Addr &corr_target, - bool actually_taken, unsigned tid); + bool actually_taken, ThreadID tid); /** * @param bp_history Pointer to the history object. The predictor @@ -178,8 +178,8 @@ class BPredUnit * information needed to update the predictor, BTB, and RAS. */ PredictorHistory(const InstSeqNum &seq_num, const Addr &inst_PC, - const bool pred_taken, void *bp_history, - const unsigned _tid) + 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) @@ -198,7 +198,7 @@ class BPredUnit unsigned RASIndex; /** The thread id. */ - unsigned tid; + ThreadID tid; /** Whether or not it was predicted taken. */ bool predTaken; diff --git a/src/cpu/inorder/resources/branch_predictor.cc b/src/cpu/inorder/resources/branch_predictor.cc index d8c0730af..905de0794 100644 --- a/src/cpu/inorder/resources/branch_predictor.cc +++ b/src/cpu/inorder/resources/branch_predictor.cc @@ -65,7 +65,7 @@ BranchPredictor::execute(int slot_num) ResourceRequest* bpred_req = reqMap[slot_num]; DynInstPtr inst = bpred_req->inst; - int tid = inst->readTid(); + ThreadID tid = inst->readTid(); int seq_num = inst->seqNum; //int stage_num = bpred_req->getStageNum(); @@ -136,14 +136,14 @@ BranchPredictor::execute(int slot_num) void BranchPredictor::squash(DynInstPtr inst, int squash_stage, - InstSeqNum squash_seq_num, unsigned tid) + InstSeqNum squash_seq_num, ThreadID tid) { DPRINTF(InOrderBPred, "Squashing...\n"); branchPred.squash(squash_seq_num, tid); } void -BranchPredictor::instGraduated(InstSeqNum seq_num,unsigned tid) +BranchPredictor::instGraduated(InstSeqNum seq_num, ThreadID tid) { branchPred.update(seq_num, tid); } diff --git a/src/cpu/inorder/resources/branch_predictor.hh b/src/cpu/inorder/resources/branch_predictor.hh index 47053910d..7d0b3348a 100644 --- a/src/cpu/inorder/resources/branch_predictor.hh +++ b/src/cpu/inorder/resources/branch_predictor.hh @@ -61,9 +61,9 @@ class BranchPredictor : public Resource { virtual void execute(int slot_num); virtual void squash(DynInstPtr inst, int stage_num, - InstSeqNum squash_seq_num, unsigned tid); + InstSeqNum squash_seq_num, ThreadID tid); - virtual void instGraduated(InstSeqNum seq_num,unsigned tid); + virtual void instGraduated(InstSeqNum seq_num, ThreadID tid); protected: /** List of instructions this resource is currently diff --git a/src/cpu/inorder/resources/cache_unit.cc b/src/cpu/inorder/resources/cache_unit.cc index 5d5d4d45d..5677810f6 100644 --- a/src/cpu/inorder/resources/cache_unit.cc +++ b/src/cpu/inorder/resources/cache_unit.cc @@ -258,17 +258,10 @@ Fault CacheUnit::doTLBAccess(DynInstPtr inst, CacheReqPtr cache_req, int acc_size, int flags, TheISA::TLB::Mode tlb_mode) { - int tid; - int seq_num; - Addr aligned_addr; - unsigned stage_num; - unsigned slot_idx; - - tid = inst->readTid(); - seq_num = inst->seqNum; - aligned_addr = inst->getMemAddr(); - stage_num = cache_req->getStageNum(); - slot_idx = cache_req->getSlot(); + ThreadID tid = inst->readTid(); + Addr aligned_addr = inst->getMemAddr(); + unsigned stage_num = cache_req->getStageNum(); + unsigned slot_idx = cache_req->getSlot(); if (tlb_mode == TheISA::TLB::Execute) { inst->fetchMemReq = new Request(inst->readTid(), aligned_addr, @@ -290,7 +283,7 @@ CacheUnit::doTLBAccess(DynInstPtr inst, CacheReqPtr cache_req, int acc_size, if (cache_req->fault != NoFault) { DPRINTF(InOrderTLB, "[tid:%i]: %s encountered while translating " "addr:%08p for [sn:%i].\n", tid, cache_req->fault->name(), - cache_req->memReq->getVaddr(), seq_num); + cache_req->memReq->getVaddr(), inst->seqNum); cpu->pipelineStage[stage_num]->setResStall(cache_req, tid); @@ -303,7 +296,7 @@ CacheUnit::doTLBAccess(DynInstPtr inst, CacheReqPtr cache_req, int acc_size, cpu->trap(cache_req->fault, tid); } else { DPRINTF(InOrderTLB, "[tid:%i]: [sn:%i] virt. addr %08p translated " - "to phys. addr:%08p.\n", tid, seq_num, + "to phys. addr:%08p.\n", tid, inst->seqNum, cache_req->memReq->getVaddr(), cache_req->memReq->getPaddr()); } @@ -361,11 +354,11 @@ CacheUnit::execute(int slot_num) assert(cache_req); DynInstPtr inst = cache_req->inst; - int tid; - int seq_num; +#if TRACING_ON + ThreadID tid = inst->readTid(); + int seq_num = inst->seqNum; +#endif - tid = inst->readTid(); - seq_num = inst->seqNum; cache_req->fault = NoFault; switch (cache_req->cmd) @@ -381,8 +374,8 @@ CacheUnit::execute(int slot_num) if (cache_req->fault == NoFault) { DPRINTF(InOrderCachePort, - "[tid:%u]: Initiating fetch access to %s for addr. %08p\n", - tid, name(), cache_req->inst->getMemAddr()); + "[tid:%u]: Initiating fetch access to %s for addr. %08p\n", + tid, name(), cache_req->inst->getMemAddr()); cache_req->reqData = new uint8_t[acc_size]; @@ -499,9 +492,9 @@ Fault CacheUnit::doCacheAccess(DynInstPtr inst, uint64_t *write_res) { Fault fault = NoFault; - int tid = 0; - - tid = inst->readTid(); +#if TRACING_ON + ThreadID tid = inst->readTid(); +#endif CacheReqPtr cache_req = dynamic_cast(reqMap[inst->getCurResSlot()]); @@ -627,10 +620,7 @@ CacheUnit::processCacheCompletion(PacketPtr pkt) // Get resource request info unsigned stage_num = cache_req->getStageNum(); DynInstPtr inst = cache_req->inst; - unsigned tid; - - - tid = cache_req->inst->readTid(); + ThreadID tid = cache_req->inst->readTid(); if (!cache_req->isSquashed()) { if (inst->resSched.top()->cmd == CompleteFetch) { @@ -752,7 +742,7 @@ CacheUnitEvent::process() { DynInstPtr inst = resource->reqMap[slotIdx]->inst; int stage_num = resource->reqMap[slotIdx]->getStageNum(); - int tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; CacheReqPtr req_ptr = dynamic_cast(resource->reqMap[slotIdx]); DPRINTF(InOrderTLB, "Waking up from TLB Miss caused by [sn:%i].\n", @@ -774,7 +764,7 @@ CacheUnitEvent::process() void CacheUnit::squash(DynInstPtr inst, int stage_num, - InstSeqNum squash_seq_num, unsigned tid) + InstSeqNum squash_seq_num, ThreadID tid) { vector slot_remove_list; diff --git a/src/cpu/inorder/resources/cache_unit.hh b/src/cpu/inorder/resources/cache_unit.hh index aba5a1b0c..8946ad5d3 100644 --- a/src/cpu/inorder/resources/cache_unit.hh +++ b/src/cpu/inorder/resources/cache_unit.hh @@ -144,7 +144,7 @@ class CacheUnit : public Resource void execute(int slot_num); void squash(DynInstPtr inst, int stage_num, - InstSeqNum squash_seq_num, unsigned tid); + InstSeqNum squash_seq_num, ThreadID tid); /** Processes cache completion event. */ void processCacheCompletion(PacketPtr pkt); diff --git a/src/cpu/inorder/resources/decode_unit.cc b/src/cpu/inorder/resources/decode_unit.cc index d95b1d4bb..033c318f2 100644 --- a/src/cpu/inorder/resources/decode_unit.cc +++ b/src/cpu/inorder/resources/decode_unit.cc @@ -39,7 +39,7 @@ DecodeUnit::DecodeUnit(std::string res_name, int res_id, int res_width, int res_latency, InOrderCPU *_cpu, ThePipeline::Params *params) : Resource(res_name, res_id, res_width, res_latency, _cpu) { - for (int tid = 0; tid < MaxThreads; tid++) { + for (ThreadID tid = 0; tid < MaxThreads; tid++) { regDepMap[tid] = &cpu->archRegDepMap[tid]; } } @@ -50,10 +50,8 @@ DecodeUnit::execute(int slot_num) ResourceRequest* decode_req = reqMap[slot_num]; DynInstPtr inst = reqMap[slot_num]->inst; Fault fault = reqMap[slot_num]->fault; - int tid, seq_num; + ThreadID tid = inst->readTid(); - tid = inst->readTid(); - seq_num = inst->seqNum; decode_req->fault = NoFault; switch (decode_req->cmd) @@ -63,13 +61,17 @@ DecodeUnit::execute(int slot_num) bool done_sked = ThePipeline::createBackEndSchedule(inst); if (done_sked) { - DPRINTF(InOrderDecode, "[tid:%i]: Setting Destination Register(s) for [sn:%i].\n", - tid, seq_num); + DPRINTF(InOrderDecode, + "[tid:%i]: Setting Destination Register(s) for [sn:%i].\n", + tid, inst->seqNum); regDepMap[tid]->insert(inst); decode_req->done(); } else { - DPRINTF(Resource,"[tid:%i] Static Inst not available to decode. Unable to create " - "schedule for instruction [sn:%i] \n", tid, inst->seqNum); + DPRINTF(Resource, + "[tid:%i] Static Inst not available to decode.\n", tid); + DPRINTF(Resource, + "Unable to create schedule for instruction [sn:%i] \n", + inst->seqNum); DPRINTF(InOrderStall, "STALL: \n"); decode_req->done(false); } @@ -83,9 +85,11 @@ DecodeUnit::execute(int slot_num) void -DecodeUnit::squash(DynInstPtr inst, int stage_num, InstSeqNum squash_seq_num, unsigned tid) +DecodeUnit::squash(DynInstPtr inst, int stage_num, InstSeqNum squash_seq_num, + ThreadID tid) { - DPRINTF(InOrderDecode, "[tid:%i]: Updating due to squash from stage %i after [sn:%i].\n", + DPRINTF(InOrderDecode, + "[tid:%i]: Updating due to squash from stage %i after [sn:%i].\n", tid, stage_num, squash_seq_num); //cpu->removeInstsUntil(squash_seq_num, tid); diff --git a/src/cpu/inorder/resources/decode_unit.hh b/src/cpu/inorder/resources/decode_unit.hh index 3813de6c4..1a700c211 100644 --- a/src/cpu/inorder/resources/decode_unit.hh +++ b/src/cpu/inorder/resources/decode_unit.hh @@ -57,7 +57,8 @@ class DecodeUnit : public Resource { virtual void execute(int slot_num); - void squash(DynInstPtr inst, int stage_num, InstSeqNum squash_seq_num, unsigned tid); + void squash(DynInstPtr inst, int stage_num, InstSeqNum squash_seq_num, + ThreadID tid); RegDepMap *regDepMap[ThePipeline::MaxThreads]; diff --git a/src/cpu/inorder/resources/execution_unit.cc b/src/cpu/inorder/resources/execution_unit.cc index 16f00b1be..c9072b5d5 100644 --- a/src/cpu/inorder/resources/execution_unit.cc +++ b/src/cpu/inorder/resources/execution_unit.cc @@ -63,7 +63,7 @@ ExecutionUnit::execute(int slot_num) ResourceRequest* exec_req = reqMap[slot_num]; DynInstPtr inst = reqMap[slot_num]->inst; Fault fault = reqMap[slot_num]->fault; - int tid = inst->readTid(); + ThreadID tid = inst->readTid(); int seq_num = inst->seqNum; exec_req->fault = NoFault; @@ -89,7 +89,7 @@ ExecutionUnit::execute(int slot_num) // that got squashed. if (inst->mispredicted()) { int stage_num = exec_req->getStageNum(); - int tid = inst->readTid(); + ThreadID tid = inst->readTid(); // If it's a branch ... if (inst->isDirectCtrl()) { diff --git a/src/cpu/inorder/resources/fetch_seq_unit.cc b/src/cpu/inorder/resources/fetch_seq_unit.cc index 69610ae58..bc809b040 100644 --- a/src/cpu/inorder/resources/fetch_seq_unit.cc +++ b/src/cpu/inorder/resources/fetch_seq_unit.cc @@ -41,7 +41,7 @@ FetchSeqUnit::FetchSeqUnit(std::string res_name, int res_id, int res_width, : Resource(res_name, res_id, res_width, res_latency, _cpu), instSize(sizeof(MachInst)) { - for (int tid = 0; tid < ThePipeline::MaxThreads; tid++) { + for (ThreadID tid = 0; tid < ThePipeline::MaxThreads; tid++) { delaySlotInfo[tid].numInsts = 0; delaySlotInfo[tid].targetReady = false; @@ -68,7 +68,7 @@ FetchSeqUnit::execute(int slot_num) // for performance considerations ResourceRequest* fs_req = reqMap[slot_num]; DynInstPtr inst = fs_req->inst; - int tid = inst->readTid(); + ThreadID tid = inst->readTid(); int stage_num = fs_req->getStageNum(); int seq_num = inst->seqNum; @@ -202,7 +202,7 @@ FetchSeqUnit::execute(int slot_num) } inline void -FetchSeqUnit::squashAfterInst(DynInstPtr inst, int stage_num, unsigned tid) +FetchSeqUnit::squashAfterInst(DynInstPtr inst, int stage_num, ThreadID tid) { // Squash In Pipeline Stage cpu->pipelineStage[stage_num]->squashDueToBranch(inst, tid); @@ -216,7 +216,7 @@ FetchSeqUnit::squashAfterInst(DynInstPtr inst, int stage_num, unsigned tid) } void FetchSeqUnit::squash(DynInstPtr inst, int squash_stage, - InstSeqNum squash_seq_num, unsigned tid) + InstSeqNum squash_seq_num, ThreadID tid) { DPRINTF(InOrderFetchSeq, "[tid:%i]: Updating due to squash from stage %i.\n", tid, squash_stage); @@ -302,7 +302,7 @@ FetchSeqUnit::FetchSeqEvent::process() void -FetchSeqUnit::activateThread(unsigned tid) +FetchSeqUnit::activateThread(ThreadID tid) { pcValid[tid] = true; @@ -317,7 +317,7 @@ FetchSeqUnit::activateThread(unsigned tid) } void -FetchSeqUnit::deactivateThread(unsigned tid) +FetchSeqUnit::deactivateThread(ThreadID tid) { delaySlotInfo[tid].numInsts = 0; delaySlotInfo[tid].targetReady = false; @@ -328,7 +328,7 @@ FetchSeqUnit::deactivateThread(unsigned tid) squashSeqNum[tid] = (InstSeqNum)-1; lastSquashCycle[tid] = 0; - std::list::iterator thread_it = find(cpu->fetchPriorityList.begin(), + list::iterator thread_it = find(cpu->fetchPriorityList.begin(), cpu->fetchPriorityList.end(), tid); diff --git a/src/cpu/inorder/resources/fetch_seq_unit.hh b/src/cpu/inorder/resources/fetch_seq_unit.hh index 1885d1f11..3e18d47cb 100644 --- a/src/cpu/inorder/resources/fetch_seq_unit.hh +++ b/src/cpu/inorder/resources/fetch_seq_unit.hh @@ -56,8 +56,8 @@ class FetchSeqUnit : public Resource { virtual ~FetchSeqUnit() {} virtual void init(); - virtual void activateThread(unsigned tid); - virtual void deactivateThread(unsigned tid); + virtual void activateThread(ThreadID tid); + virtual void deactivateThread(ThreadID tid); virtual void execute(int slot_num); /** Override default Resource squash sequence. This actually, @@ -65,10 +65,10 @@ class FetchSeqUnit : public Resource { * info */ virtual void squash(DynInstPtr inst, int squash_stage, - InstSeqNum squash_seq_num, unsigned tid); + InstSeqNum squash_seq_num, ThreadID tid); - inline void squashAfterInst(DynInstPtr inst, int stage_num, unsigned tid); + inline void squashAfterInst(DynInstPtr inst, int stage_num, ThreadID tid); protected: unsigned instSize; diff --git a/src/cpu/inorder/resources/graduation_unit.cc b/src/cpu/inorder/resources/graduation_unit.cc index 569401e4f..2d7cd5c8c 100644 --- a/src/cpu/inorder/resources/graduation_unit.cc +++ b/src/cpu/inorder/resources/graduation_unit.cc @@ -39,7 +39,7 @@ GraduationUnit::GraduationUnit(std::string res_name, int res_id, int res_width, lastCycleGrad(0), numCycleGrad(0) { - for (int tid = 0; tid < ThePipeline::MaxThreads; tid++) { + for (ThreadID tid = 0; tid < ThePipeline::MaxThreads; tid++) { nonSpecInstActive[tid] = &cpu->nonSpecInstActive[tid]; nonSpecSeqNum[tid] = &cpu->nonSpecSeqNum[tid]; } @@ -51,10 +51,7 @@ GraduationUnit::execute(int slot_num) ResourceRequest* grad_req = reqMap[slot_num]; DynInstPtr inst = reqMap[slot_num]->inst; Fault fault = reqMap[slot_num]->fault; - int tid, seq_num; - - tid = inst->readTid(); - seq_num = inst->seqNum; + ThreadID tid = inst->readTid(); int stage_num = inst->resSched.top()->stageNum; grad_req->fault = NoFault; @@ -70,15 +67,17 @@ GraduationUnit::execute(int slot_num) lastCycleGrad = curTick; numCycleGrad = 0; } else if (numCycleGrad > width) { - DPRINTF(InOrderGraduation, "Graduation bandwidth reached for this cycle.\n"); + DPRINTF(InOrderGraduation, + "Graduation bandwidth reached for this cycle.\n"); return; } // Make sure this is the last thing on the resource schedule assert(inst->resSched.size() == 1); - DPRINTF(InOrderGraduation, "[tid:%i] Graduating instruction [sn:%i].\n", - tid, seq_num); + DPRINTF(InOrderGraduation, + "[tid:%i] Graduating instruction [sn:%i].\n", + tid, inst->seqNum); DPRINTF(RefCount, "Refcount = %i.\n", 0/*inst->curCount()*/); @@ -87,8 +86,9 @@ GraduationUnit::execute(int slot_num) // @TODO: Fix this functionality. Probably too conservative. if (inst->isNonSpeculative()) { *nonSpecInstActive[tid] = false; - DPRINTF(InOrderGraduation, "[tid:%i] Non-speculative instruction [sn:%i] has graduated.\n", - tid, seq_num); + DPRINTF(InOrderGraduation, + "[tid:%i] Non-speculative inst [sn:%i] graduated\n", + tid, inst->seqNum); } if (inst->traceData) { diff --git a/src/cpu/inorder/resources/inst_buffer.cc b/src/cpu/inorder/resources/inst_buffer.cc index fafff1fa7..21df1d053 100644 --- a/src/cpu/inorder/resources/inst_buffer.cc +++ b/src/cpu/inorder/resources/inst_buffer.cc @@ -60,11 +60,9 @@ InstBuffer::execute(int slot_idx) { ResReqPtr ib_req = reqMap[slot_idx]; DynInstPtr inst = ib_req->inst; - int tid, seq_num, stage_num; + ThreadID tid = inst->readTid(); + int stage_num = ib_req->getStageNum(); - tid = inst->readTid(); - seq_num = inst->seqNum; - stage_num = ib_req->getStageNum(); ib_req->fault = NoFault; switch (ib_req->cmd) @@ -121,12 +119,12 @@ InstBuffer::execute(int slot_idx) if (instList.size() < width) { DPRINTF(InOrderInstBuffer, "[tid:%i]: Inserting [sn:%i] into buffer.\n", - tid, seq_num); + tid, inst->seqNum); insert(inst); inserted = true; } else { DPRINTF(InOrderInstBuffer, "[tid:%i]: Denying [sn:%i] request because " - "buffer is full.\n", tid, seq_num); + "buffer is full.\n", tid, inst->seqNum); std::list::iterator list_it = instList.begin(); @@ -145,7 +143,7 @@ InstBuffer::execute(int slot_idx) case RemoveInst: { DPRINTF(InOrderInstBuffer, "[tid:%i]: Removing [sn:%i] from buffer.\n", - tid, seq_num); + tid, inst->seqNum); remove(inst); ib_req->done(); } @@ -180,20 +178,20 @@ InstBuffer::remove(DynInstPtr inst) } void -InstBuffer::pop(unsigned tid) +InstBuffer::pop(ThreadID tid) { instList.pop_front(); } ThePipeline::DynInstPtr -InstBuffer::top(unsigned tid) +InstBuffer::top(ThreadID tid) { return instList.front(); } void InstBuffer::squash(DynInstPtr inst, int stage_num, - InstSeqNum squash_seq_num, unsigned tid) + InstSeqNum squash_seq_num, ThreadID tid) { queue::iterator> remove_list; list::iterator list_it = instList.begin(); diff --git a/src/cpu/inorder/resources/inst_buffer.hh b/src/cpu/inorder/resources/inst_buffer.hh index baadd42ff..7342df66f 100644 --- a/src/cpu/inorder/resources/inst_buffer.hh +++ b/src/cpu/inorder/resources/inst_buffer.hh @@ -67,12 +67,12 @@ class InstBuffer : public Resource { virtual void remove(DynInstPtr inst); - virtual void pop(unsigned tid); + virtual void pop(ThreadID tid); - virtual DynInstPtr top(unsigned tid); + virtual DynInstPtr top(ThreadID tid); virtual void squash(DynInstPtr inst, int stage_num, - InstSeqNum squash_seq_num, unsigned tid); + InstSeqNum squash_seq_num, ThreadID tid); protected: /** List of instructions this resource is currently * processing. diff --git a/src/cpu/inorder/resources/inst_buffer_new.cc b/src/cpu/inorder/resources/inst_buffer_new.cc index 7e2c98837..cc534ef3e 100644 --- a/src/cpu/inorder/resources/inst_buffer_new.cc +++ b/src/cpu/inorder/resources/inst_buffer_new.cc @@ -67,7 +67,7 @@ InstBuffer::execute(int slot_idx) assert(ib_req); DynInstPtr inst = ib_req->inst; - int tid = inst->readTid(); + ThreadID tid = inst->readTid(); int seq_num = inst->seqNum; ib_req->fault = NoFault; @@ -128,7 +128,7 @@ InstBuffer::top() { return instList.front(); } void -InstBuffer::squash(InstSeqNum squash_seq_num, unsigned tid) +InstBuffer::squash(InstSeqNum squash_seq_num, ThreadID tid) { list::iterator list_it = instList.begin(); list::iterator list_end = instList.end(); diff --git a/src/cpu/inorder/resources/inst_buffer_new.hh b/src/cpu/inorder/resources/inst_buffer_new.hh index e374fa109..b1d5a7b09 100644 --- a/src/cpu/inorder/resources/inst_buffer_new.hh +++ b/src/cpu/inorder/resources/inst_buffer_new.hh @@ -71,7 +71,7 @@ class InstBuffer : public Resource { virtual DynInstPtr top(); - virtual void squash(InstSeqNum squash_seq_num, unsigned tid); + virtual void squash(InstSeqNum squash_seq_num, ThreadID tid); protected: /** List of instructions this resource is currently diff --git a/src/cpu/inorder/resources/mult_div_unit.cc b/src/cpu/inorder/resources/mult_div_unit.cc index df9d4c293..7592c0260 100644 --- a/src/cpu/inorder/resources/mult_div_unit.cc +++ b/src/cpu/inorder/resources/mult_div_unit.cc @@ -174,7 +174,7 @@ MultDivUnit::execute(int slot_num) DynInstPtr inst = reqMap[slot_num]->inst; Fault fault = reqMap[slot_num]->fault; - //int tid = inst->readTid(); + //ThreadID tid = inst->readTid(); //int seq_num = inst->seqNum; switch (mult_div_req->cmd) @@ -248,7 +248,7 @@ MultDivUnit::exeMulDiv(int slot_num) ResourceRequest* mult_div_req = reqMap[slot_num]; DynInstPtr inst = reqMap[slot_num]->inst; Fault fault = reqMap[slot_num]->fault; - int tid = inst->readTid(); + ThreadID tid = inst->readTid(); int seq_num = inst->seqNum; fault = inst->execute(); diff --git a/src/cpu/inorder/resources/tlb_unit.cc b/src/cpu/inorder/resources/tlb_unit.cc index 1ce8ff8c2..95bade36a 100644 --- a/src/cpu/inorder/resources/tlb_unit.cc +++ b/src/cpu/inorder/resources/tlb_unit.cc @@ -101,11 +101,9 @@ TLBUnit::execute(int slot_idx) assert(tlb_req != 0x0); DynInstPtr inst = tlb_req->inst; - int tid, seq_num, stage_num; - - tid = inst->readTid(); - seq_num = inst->seqNum; - stage_num = tlb_req->getStageNum(); + ThreadID tid = inst->readTid(); + int seq_num = inst->seqNum; + int stage_num = tlb_req->getStageNum(); tlb_req->fault = NoFault; @@ -202,7 +200,7 @@ TLBUnitEvent::process() { DynInstPtr inst = resource->reqMap[slotIdx]->inst; int stage_num = resource->reqMap[slotIdx]->getStageNum(); - int tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; DPRINTF(InOrderTLB, "Waking up from TLB Miss caused by [sn:%i].\n", inst->seqNum); @@ -224,7 +222,7 @@ TLBUnitEvent::process() void TLBUnit::squash(DynInstPtr inst, int stage_num, - InstSeqNum squash_seq_num, unsigned tid) + InstSeqNum squash_seq_num, ThreadID tid) { //@TODO: Figure out a way to consolidate common parts // of this squash code diff --git a/src/cpu/inorder/resources/tlb_unit.hh b/src/cpu/inorder/resources/tlb_unit.hh index 8f0291b48..1c08bd822 100644 --- a/src/cpu/inorder/resources/tlb_unit.hh +++ b/src/cpu/inorder/resources/tlb_unit.hh @@ -67,7 +67,8 @@ class TLBUnit : public Resource virtual void execute(int slot_num); - void squash(DynInstPtr inst, int stage_num, InstSeqNum squash_seq_num, unsigned tid); + void squash(DynInstPtr inst, int stage_num, InstSeqNum squash_seq_num, + ThreadID tid); bool tlbBlocked[ThePipeline::MaxThreads]; diff --git a/src/cpu/inorder/resources/use_def.cc b/src/cpu/inorder/resources/use_def.cc index 53145640e..b30a3a1bf 100644 --- a/src/cpu/inorder/resources/use_def.cc +++ b/src/cpu/inorder/resources/use_def.cc @@ -45,7 +45,7 @@ UseDefUnit::UseDefUnit(string res_name, int res_id, int res_width, : Resource(res_name, res_id, res_width, res_latency, _cpu), maxSeqNum((InstSeqNum)-1) { - for (int tid = 0; tid < ThePipeline::MaxThreads; tid++) { + for (ThreadID tid = 0; tid < ThePipeline::MaxThreads; tid++) { nonSpecInstActive[tid] = &cpu->nonSpecInstActive[tid]; nonSpecSeqNum[tid] = &cpu->nonSpecSeqNum[tid]; @@ -99,7 +99,7 @@ UseDefUnit::execute(int slot_idx) assert(ud_req); DynInstPtr inst = ud_req->inst; - int tid = inst->readTid(); + ThreadID tid = inst->readTid(); int seq_num = inst->seqNum; int ud_idx = ud_req->useDefIdx; @@ -306,7 +306,8 @@ UseDefUnit::execute(int slot_idx) } void -UseDefUnit::squash(DynInstPtr inst, int stage_num, InstSeqNum squash_seq_num, unsigned tid) +UseDefUnit::squash(DynInstPtr inst, int stage_num, InstSeqNum squash_seq_num, + ThreadID tid) { DPRINTF(InOrderUseDef, "[tid:%i]: Updating Due To Squash After [sn:%i].\n", tid, squash_seq_num); diff --git a/src/cpu/inorder/resources/use_def.hh b/src/cpu/inorder/resources/use_def.hh index 51ec2c3f2..6c76d8ab5 100644 --- a/src/cpu/inorder/resources/use_def.hh +++ b/src/cpu/inorder/resources/use_def.hh @@ -65,7 +65,8 @@ class UseDefUnit : public Resource { virtual void execute(int slot_num); - virtual void squash(DynInstPtr inst, int stage_num, InstSeqNum squash_seq_num, unsigned tid); + virtual void squash(DynInstPtr inst, int stage_num, + InstSeqNum squash_seq_num, ThreadID tid); const InstSeqNum maxSeqNum; diff --git a/src/cpu/inorder/thread_context.cc b/src/cpu/inorder/thread_context.cc index 0cac51559..a1e9b5948 100644 --- a/src/cpu/inorder/thread_context.cc +++ b/src/cpu/inorder/thread_context.cc @@ -175,7 +175,7 @@ InOrderThreadContext::readFloatRegBits(int reg_idx) } uint64_t -InOrderThreadContext::readRegOtherThread(int reg_idx, unsigned tid) +InOrderThreadContext::readRegOtherThread(int reg_idx, ThreadID tid) { return cpu->readRegOtherThread(reg_idx, tid); } @@ -212,7 +212,8 @@ InOrderThreadContext::setFloatRegBits(int reg_idx, FloatRegBits val) } void -InOrderThreadContext::setRegOtherThread(int misc_reg, const MiscReg &val, unsigned tid) +InOrderThreadContext::setRegOtherThread(int misc_reg, const MiscReg &val, + ThreadID tid) { cpu->setRegOtherThread(misc_reg, val, tid); } diff --git a/src/cpu/inorder/thread_context.hh b/src/cpu/inorder/thread_context.hh index cde377dfc..f3cf3ec44 100644 --- a/src/cpu/inorder/thread_context.hh +++ b/src/cpu/inorder/thread_context.hh @@ -160,7 +160,7 @@ class InOrderThreadContext : public ThreadContext virtual FloatRegBits readFloatRegBits(int reg_idx); - virtual uint64_t readRegOtherThread(int misc_reg, unsigned tid); + virtual uint64_t readRegOtherThread(int misc_reg, ThreadID tid); /** Sets an integer register to a value. */ virtual void setIntReg(int reg_idx, uint64_t val); @@ -173,7 +173,8 @@ class InOrderThreadContext : public ThreadContext virtual void setFloatRegBits(int reg_idx, FloatRegBits val); - virtual void setRegOtherThread(int misc_reg, const MiscReg &val, unsigned tid); + virtual void setRegOtherThread(int misc_reg, const MiscReg &val, + ThreadID tid); /** Reads this thread's PC. */ virtual uint64_t readPC() diff --git a/src/cpu/inorder/thread_state.hh b/src/cpu/inorder/thread_state.hh index eb4fe40b2..803659487 100644 --- a/src/cpu/inorder/thread_state.hh +++ b/src/cpu/inorder/thread_state.hh @@ -67,8 +67,10 @@ class InOrderThreadState : public ThreadState { bool trapPending; - InOrderThreadState(InOrderCPU *_cpu, int _thread_num, Process *_process, int _asid) - : ThreadState(reinterpret_cast(_cpu), 0/*_thread_num*/, _process, 0/*_asid*/), + InOrderThreadState(InOrderCPU *_cpu, ThreadID _thread_num, + Process *_process, int _asid) + : ThreadState(reinterpret_cast(_cpu), 0/*_thread_num*/, + _process, 0/*_asid*/), cpu(_cpu), inSyscall(0), trapPending(0) { } diff --git a/src/cpu/o3/bpred_unit.hh b/src/cpu/o3/bpred_unit.hh index 4875c03d8..49cb4233b 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, unsigned tid); + bool predict(DynInstPtr &inst, Addr &PC, ThreadID tid); // @todo: Rename this function. void BPUncond(void * &bp_history); @@ -99,7 +99,7 @@ class BPredUnit * @param done_sn The sequence number to commit any older updates up until. * @param tid The thread id. */ - void update(const InstSeqNum &done_sn, unsigned tid); + void update(const InstSeqNum &done_sn, ThreadID tid); /** * Squashes all outstanding updates until a given sequence number. @@ -107,7 +107,7 @@ class BPredUnit * until. * @param tid The thread id. */ - void squash(const InstSeqNum &squashed_sn, unsigned tid); + void squash(const InstSeqNum &squashed_sn, ThreadID tid); /** * Squashes all outstanding updates until a given sequence number, and @@ -119,7 +119,7 @@ class BPredUnit * @param tid The thread id. */ void squash(const InstSeqNum &squashed_sn, const Addr &corr_target, - bool actually_taken, unsigned tid); + bool actually_taken, ThreadID tid); /** * @param bp_history Pointer to the history object. The predictor @@ -179,8 +179,8 @@ class BPredUnit * information needed to update the predictor, BTB, and RAS. */ PredictorHistory(const InstSeqNum &seq_num, const Addr &inst_PC, - const bool pred_taken, void *bp_history, - const unsigned _tid) + 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) @@ -203,7 +203,7 @@ class BPredUnit unsigned RASIndex; /** The thread id. */ - unsigned tid; + ThreadID tid; /** Whether or not it was predicted taken. */ bool predTaken; diff --git a/src/cpu/o3/bpred_unit_impl.hh b/src/cpu/o3/bpred_unit_impl.hh index 26aba986d..1378ac135 100644 --- a/src/cpu/o3/bpred_unit_impl.hh +++ b/src/cpu/o3/bpred_unit_impl.hh @@ -144,7 +144,7 @@ BPredUnit::takeOverFrom() template bool -BPredUnit::predict(DynInstPtr &inst, Addr &PC, unsigned tid) +BPredUnit::predict(DynInstPtr &inst, Addr &PC, ThreadID tid) { // See if branch predictor predicts taken. // If so, get its target addr either from the BTB or the RAS. @@ -254,7 +254,7 @@ BPredUnit::predict(DynInstPtr &inst, Addr &PC, unsigned tid) template void -BPredUnit::update(const InstSeqNum &done_sn, unsigned tid) +BPredUnit::update(const InstSeqNum &done_sn, ThreadID tid) { DPRINTF(Fetch, "BranchPred: [tid:%i]: Committing branches until " "[sn:%lli].\n", tid, done_sn); @@ -272,7 +272,7 @@ BPredUnit::update(const InstSeqNum &done_sn, unsigned tid) template void -BPredUnit::squash(const InstSeqNum &squashed_sn, unsigned tid) +BPredUnit::squash(const InstSeqNum &squashed_sn, ThreadID tid) { History &pred_hist = predHist[tid]; @@ -311,8 +311,8 @@ template void BPredUnit::squash(const InstSeqNum &squashed_sn, const Addr &corr_target, - const bool actually_taken, - unsigned tid) + bool actually_taken, + ThreadID tid) { // Now that we know that a branch was mispredicted, we need to undo // all the branches that have been seen up until this branch and diff --git a/src/cpu/o3/btb.hh b/src/cpu/o3/btb.hh index 38ecabc46..6557522e0 100644 --- a/src/cpu/o3/btb.hh +++ b/src/cpu/o3/btb.hh @@ -51,7 +51,7 @@ class DefaultBTB Addr target; /** The entry's thread id. */ - unsigned tid; + ThreadID tid; /** Whether or not the entry is valid. */ bool valid; @@ -74,14 +74,14 @@ class DefaultBTB * @param tid The thread id. * @return Returns the target of the branch. */ - Addr lookup(const Addr &inst_PC, unsigned tid); + Addr lookup(const Addr &inst_PC, 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, unsigned tid); + bool valid(const Addr &inst_PC, ThreadID tid); /** Updates the BTB with the target of a branch. * @param inst_PC The address of the branch being updated. @@ -89,7 +89,7 @@ class DefaultBTB * @param tid The thread id. */ void update(const Addr &inst_PC, const Addr &target_PC, - unsigned tid); + ThreadID tid); private: /** Returns the index into the BTB, based on the branch's PC. diff --git a/src/cpu/o3/commit.hh b/src/cpu/o3/commit.hh index f21c14569..d93b85984 100644 --- a/src/cpu/o3/commit.hh +++ b/src/cpu/o3/commit.hh @@ -92,10 +92,10 @@ class DefaultCommit class TrapEvent : public Event { private: DefaultCommit *commit; - unsigned tid; + ThreadID tid; public: - TrapEvent(DefaultCommit *_commit, unsigned _tid); + TrapEvent(DefaultCommit *_commit, ThreadID _tid); void process(); const char *description() const; @@ -172,7 +172,7 @@ class DefaultCommit IEW *iewStage; /** Sets pointer to list of active threads. */ - void setActiveThreads(std::list *at_ptr); + void setActiveThreads(std::list *at_ptr); /** Sets pointer to the commited state rename map. */ void setRenameMap(RenameMap rm_ptr[Impl::MaxThreads]); @@ -204,15 +204,15 @@ class DefaultCommit void commit(); /** Returns the number of free ROB entries for a specific thread. */ - unsigned numROBFreeEntries(unsigned tid); + size_t numROBFreeEntries(ThreadID tid); /** Generates an event to schedule a squash due to a trap. */ - void generateTrapEvent(unsigned tid); + void generateTrapEvent(ThreadID tid); /** Records that commit needs to initiate a squash due to an * external state update through the TC. */ - void generateTCEvent(unsigned tid); + void generateTCEvent(ThreadID tid); private: /** Updates the overall status of commit with the nextStatus, and @@ -237,13 +237,13 @@ class DefaultCommit bool changedROBEntries(); /** Squashes all in flight instructions. */ - void squashAll(unsigned tid); + void squashAll(ThreadID tid); /** Handles squashing due to a trap. */ - void squashFromTrap(unsigned tid); + void squashFromTrap(ThreadID tid); /** Handles squashing due to an TC write. */ - void squashFromTC(unsigned tid); + void squashFromTC(ThreadID tid); #if FULL_SYSTEM /** Handles processing an interrupt. */ @@ -268,13 +268,13 @@ class DefaultCommit void markCompletedInsts(); /** Gets the thread to commit, based on the SMT policy. */ - int getCommittingThread(); + ThreadID getCommittingThread(); /** Returns the thread ID to use based on a round robin policy. */ - int roundRobin(); + ThreadID roundRobin(); /** Returns the thread ID to use based on an oldest instruction policy. */ - int oldestReady(); + ThreadID oldestReady(); public: /** Returns the PC of the head instruction of the ROB. @@ -283,34 +283,34 @@ class DefaultCommit Addr readPC() { return PC[0]; } /** Returns the PC of a specific thread. */ - Addr readPC(unsigned tid) { return PC[tid]; } + Addr readPC(ThreadID tid) { return PC[tid]; } /** Sets the PC of a specific thread. */ - void setPC(Addr val, unsigned tid) { PC[tid] = val; } + void setPC(Addr val, ThreadID tid) { PC[tid] = val; } /** Reads the micro PC of a specific thread. */ - Addr readMicroPC(unsigned tid) { return microPC[tid]; } + Addr readMicroPC(ThreadID tid) { return microPC[tid]; } /** Sets the micro PC of a specific thread */ - void setMicroPC(Addr val, unsigned tid) { microPC[tid] = val; } + void setMicroPC(Addr val, ThreadID tid) { microPC[tid] = val; } /** Reads the next PC of a specific thread. */ - Addr readNextPC(unsigned tid) { return nextPC[tid]; } + Addr readNextPC(ThreadID tid) { return nextPC[tid]; } /** Sets the next PC of a specific thread. */ - void setNextPC(Addr val, unsigned tid) { nextPC[tid] = val; } + void setNextPC(Addr val, ThreadID tid) { nextPC[tid] = val; } /** Reads the next NPC of a specific thread. */ - Addr readNextNPC(unsigned tid) { return nextNPC[tid]; } + Addr readNextNPC(ThreadID tid) { return nextNPC[tid]; } /** Sets the next NPC of a specific thread. */ - void setNextNPC(Addr val, unsigned tid) { nextNPC[tid] = val; } + void setNextNPC(Addr val, ThreadID tid) { nextNPC[tid] = val; } /** Reads the micro PC of a specific thread. */ - Addr readNextMicroPC(unsigned tid) { return nextMicroPC[tid]; } + Addr readNextMicroPC(ThreadID tid) { return nextMicroPC[tid]; } /** Sets the micro PC of a specific thread */ - void setNextMicroPC(Addr val, unsigned tid) { nextMicroPC[tid] = val; } + void setNextMicroPC(Addr val, ThreadID tid) { nextMicroPC[tid] = val; } private: /** Time buffer interface. */ @@ -360,7 +360,7 @@ class DefaultCommit bool changedROBNumEntries[Impl::MaxThreads]; /** A counter of how many threads are currently squashing. */ - int squashCounter; + ThreadID squashCounter; /** Records if a thread has to squash this cycle due to a trap. */ bool trapSquash[Impl::MaxThreads]; @@ -369,7 +369,7 @@ class DefaultCommit bool tcSquash[Impl::MaxThreads]; /** Priority List used for Commit Policy */ - std::list priority_list; + std::list priority_list; /** IEW to Commit delay, in ticks. */ unsigned iewToCommitDelay; @@ -394,7 +394,7 @@ class DefaultCommit unsigned numRobs; /** Number of Active Threads */ - unsigned numThreads; + ThreadID numThreads; /** Is a drain pending. */ bool drainPending; @@ -443,7 +443,7 @@ class DefaultCommit bool checkEmptyROB[Impl::MaxThreads]; /** Pointer to the list of active threads. */ - std::list *activeThreads; + std::list *activeThreads; /** Rename map interface. */ RenameMap *renameMap[Impl::MaxThreads]; diff --git a/src/cpu/o3/commit_impl.hh b/src/cpu/o3/commit_impl.hh index 7cd88b49b..7286f1b6f 100644 --- a/src/cpu/o3/commit_impl.hh +++ b/src/cpu/o3/commit_impl.hh @@ -29,9 +29,6 @@ * Korey Sewell */ -#include "config/full_system.hh" -#include "config/use_checker.hh" - #include #include @@ -39,19 +36,22 @@ #include "base/cp_annotate.hh" #include "base/loader/symtab.hh" #include "base/timebuf.hh" +#include "config/full_system.hh" +#include "config/use_checker.hh" #include "cpu/exetrace.hh" #include "cpu/o3/commit.hh" #include "cpu/o3/thread_state.hh" +#include "params/DerivO3CPU.hh" #if USE_CHECKER #include "cpu/checker/cpu.hh" #endif -#include "params/DerivO3CPU.hh" +using namespace std; template DefaultCommit::TrapEvent::TrapEvent(DefaultCommit *_commit, - unsigned _tid) + ThreadID _tid) : Event(CPU_Tick_Pri), commit(_commit), tid(_tid) { this->setFlags(AutoDelete); @@ -105,7 +105,7 @@ DefaultCommit::DefaultCommit(O3CPU *_cpu, DerivO3CPUParams *params) commitPolicy = RoundRobin; //Set-Up Priority List - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { priority_list.push_back(tid); } @@ -119,15 +119,19 @@ DefaultCommit::DefaultCommit(O3CPU *_cpu, DerivO3CPUParams *params) "RoundRobin,OldestReady}"); } - for (int i=0; i < numThreads; i++) { - commitStatus[i] = Idle; - changedROBNumEntries[i] = false; - checkEmptyROB[i] = false; - trapInFlight[i] = false; - committedStores[i] = false; - trapSquash[i] = false; - tcSquash[i] = false; - microPC[i] = nextMicroPC[i] = PC[i] = nextPC[i] = nextNPC[i] = 0; + for (ThreadID tid = 0; tid < numThreads; tid++) { + commitStatus[tid] = Idle; + changedROBNumEntries[tid] = false; + checkEmptyROB[tid] = false; + trapInFlight[tid] = false; + committedStores[tid] = false; + trapSquash[tid] = false; + tcSquash[tid] = false; + microPC[tid] = 0; + nextMicroPC[tid] = 0; + PC[tid] = 0; + nextPC[tid] = 0; + nextNPC[tid] = 0; } #if FULL_SYSTEM interrupt = NoFault; @@ -175,49 +179,49 @@ DefaultCommit::regStats() ; statComInst - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:count") .desc("Number of instructions committed") .flags(total) ; statComSwp - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:swp_count") .desc("Number of s/w prefetches committed") .flags(total) ; statComRefs - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:refs") .desc("Number of memory references committed") .flags(total) ; statComLoads - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:loads") .desc("Number of loads committed") .flags(total) ; statComMembars - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:membars") .desc("Number of memory barriers committed") .flags(total) ; statComBranches - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:branches") .desc("Number of branches committed") .flags(total) ; commitEligible - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:bw_limited") .desc("number of insts not committed due to BW limits") .flags(total) @@ -288,7 +292,7 @@ DefaultCommit::setIEWStage(IEW *iew_stage) template void -DefaultCommit::setActiveThreads(std::list *at_ptr) +DefaultCommit::setActiveThreads(list *at_ptr) { activeThreads = at_ptr; } @@ -297,9 +301,8 @@ template void DefaultCommit::setRenameMap(RenameMap rm_ptr[]) { - for (int i=0; i < numThreads; i++) { - renameMap[i] = &rm_ptr[i]; - } + for (ThreadID tid = 0; tid < numThreads; tid++) + renameMap[tid] = &rm_ptr[tid]; } template @@ -317,10 +320,10 @@ DefaultCommit::initStage() rob->resetEntries(); // Broadcast the number of free entries. - for (int i=0; i < numThreads; i++) { - toIEW->commitInfo[i].usedROB = true; - toIEW->commitInfo[i].freeROBEntries = rob->numFreeEntries(i); - toIEW->commitInfo[i].emptyROB = true; + for (ThreadID tid = 0; tid < numThreads; tid++) { + toIEW->commitInfo[tid].usedROB = true; + toIEW->commitInfo[tid].freeROBEntries = rob->numFreeEntries(tid); + toIEW->commitInfo[tid].emptyROB = true; } // Commit must broadcast the number of free entries it has at the @@ -363,11 +366,11 @@ DefaultCommit::takeOverFrom() switchedOut = false; _status = Active; _nextStatus = Inactive; - for (int i=0; i < numThreads; i++) { - commitStatus[i] = Idle; - changedROBNumEntries[i] = false; - trapSquash[i] = false; - tcSquash[i] = false; + for (ThreadID tid = 0; tid < numThreads; tid++) { + commitStatus[tid] = Idle; + changedROBNumEntries[tid] = false; + trapSquash[tid] = false; + tcSquash[tid] = false; } squashCounter = 0; rob->takeOverFrom(); @@ -378,11 +381,11 @@ void DefaultCommit::updateStatus() { // reset ROB changed variable - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; changedROBNumEntries[tid] = false; @@ -410,11 +413,11 @@ DefaultCommit::setNextStatus() { int squashes = 0; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (commitStatus[tid] == ROBSquashing) { squashes++; @@ -434,11 +437,11 @@ template bool DefaultCommit::changedROBEntries() { - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (changedROBNumEntries[tid]) { return true; @@ -449,15 +452,15 @@ DefaultCommit::changedROBEntries() } template -unsigned -DefaultCommit::numROBFreeEntries(unsigned tid) +size_t +DefaultCommit::numROBFreeEntries(ThreadID tid) { return rob->numFreeEntries(tid); } template void -DefaultCommit::generateTrapEvent(unsigned tid) +DefaultCommit::generateTrapEvent(ThreadID tid) { DPRINTF(Commit, "Generating trap event for [tid:%i]\n", tid); @@ -469,7 +472,7 @@ DefaultCommit::generateTrapEvent(unsigned tid) template void -DefaultCommit::generateTCEvent(unsigned tid) +DefaultCommit::generateTCEvent(ThreadID tid) { assert(!trapInFlight[tid]); DPRINTF(Commit, "Generating TC squash event for [tid:%i]\n", tid); @@ -479,7 +482,7 @@ DefaultCommit::generateTCEvent(unsigned tid) template void -DefaultCommit::squashAll(unsigned tid) +DefaultCommit::squashAll(ThreadID tid) { // If we want to include the squashing instruction in the squash, // then use one older sequence number. @@ -516,7 +519,7 @@ DefaultCommit::squashAll(unsigned tid) template void -DefaultCommit::squashFromTrap(unsigned tid) +DefaultCommit::squashFromTrap(ThreadID tid) { squashAll(tid); @@ -534,7 +537,7 @@ DefaultCommit::squashFromTrap(unsigned tid) template void -DefaultCommit::squashFromTC(unsigned tid) +DefaultCommit::squashFromTC(ThreadID tid) { squashAll(tid); @@ -565,13 +568,13 @@ DefaultCommit::tick() if (activeThreads->empty()) return; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); // Check if any of the threads are done squashing. Change the // status if they are done. while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; // Clear the bit saying if the thread has committed stores // this cycle. @@ -598,7 +601,7 @@ DefaultCommit::tick() threads = activeThreads->begin(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!rob->isEmpty(tid) && rob->readHeadInst(tid)->readyToCommit()) { // The ROB has more instructions it can commit. Its next status @@ -704,11 +707,11 @@ DefaultCommit::commit() //////////////////////////////////// // Check for any possible squashes, handle them first //////////////////////////////////// - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; // Not sure which one takes priority. I think if we have // both, that's a bad sign. @@ -794,7 +797,7 @@ DefaultCommit::commit() threads = activeThreads->begin(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (changedROBNumEntries[tid]) { toIEW->commitInfo[tid].usedROB = true; @@ -855,7 +858,7 @@ DefaultCommit::commitInsts() head_inst = rob->readHeadInst(commit_thread); - int tid = head_inst->threadNumber; + ThreadID tid = head_inst->threadNumber; assert(tid == commit_thread); @@ -951,7 +954,7 @@ DefaultCommit::commitHead(DynInstPtr &head_inst, unsigned inst_num) { assert(head_inst); - int tid = head_inst->threadNumber; + ThreadID tid = head_inst->threadNumber; // If the instruction is not executed yet, then it will need extra // handling. Signal backwards that it should be executed. @@ -1147,7 +1150,7 @@ DefaultCommit::getInsts() DynInstPtr inst; inst = fromRename->insts[inst_num]; - int tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; if (!inst->isSquashed() && commitStatus[tid] != ROBSquashing && @@ -1220,11 +1223,11 @@ template bool DefaultCommit::robDoneSquashing() { - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!rob->isDoneSquashing(tid)) return false; @@ -1237,40 +1240,40 @@ template void DefaultCommit::updateComInstStats(DynInstPtr &inst) { - unsigned thread = inst->threadNumber; + ThreadID tid = inst->threadNumber; // // Pick off the software prefetches // #ifdef TARGET_ALPHA if (inst->isDataPrefetch()) { - statComSwp[thread]++; + statComSwp[tid]++; } else { - statComInst[thread]++; + statComInst[tid]++; } #else - statComInst[thread]++; + statComInst[tid]++; #endif // // Control Instructions // if (inst->isControl()) - statComBranches[thread]++; + statComBranches[tid]++; // // Memory references // if (inst->isMemRef()) { - statComRefs[thread]++; + statComRefs[tid]++; if (inst->isLoad()) { - statComLoads[thread]++; + statComLoads[tid]++; } } if (inst->isMemBarrier()) { - statComMembars[thread]++; + statComMembars[tid]++; } } @@ -1280,7 +1283,7 @@ DefaultCommit::updateComInstStats(DynInstPtr &inst) // // //////////////////////////////////////// template -int +ThreadID DefaultCommit::getCommittingThread() { if (numThreads > 1) { @@ -1299,31 +1302,31 @@ DefaultCommit::getCommittingThread() return oldestReady(); default: - return -1; + return InvalidThreadID; } } else { assert(!activeThreads->empty()); - int tid = activeThreads->front(); + ThreadID tid = activeThreads->front(); if (commitStatus[tid] == Running || commitStatus[tid] == Idle || commitStatus[tid] == FetchTrapPending) { return tid; } else { - return -1; + return InvalidThreadID; } } } template -int +ThreadID DefaultCommit::roundRobin() { - std::list::iterator pri_iter = priority_list.begin(); - std::list::iterator end = priority_list.end(); + list::iterator pri_iter = priority_list.begin(); + list::iterator end = priority_list.end(); while (pri_iter != end) { - unsigned tid = *pri_iter; + ThreadID tid = *pri_iter; if (commitStatus[tid] == Running || commitStatus[tid] == Idle || @@ -1340,21 +1343,21 @@ DefaultCommit::roundRobin() pri_iter++; } - return -1; + return InvalidThreadID; } template -int +ThreadID DefaultCommit::oldestReady() { unsigned oldest = 0; bool first = true; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!rob->isEmpty(tid) && (commitStatus[tid] == Running || @@ -1378,6 +1381,6 @@ DefaultCommit::oldestReady() if (!first) { return oldest; } else { - return -1; + return InvalidThreadID; } } diff --git a/src/cpu/o3/cpu.cc b/src/cpu/o3/cpu.cc index 038e5daaa..621b6c1b9 100644 --- a/src/cpu/o3/cpu.cc +++ b/src/cpu/o3/cpu.cc @@ -60,6 +60,7 @@ class BaseCPUParams; using namespace TheISA; +using namespace std; BaseO3CPU::BaseO3CPU(BaseCPUParams *params) : BaseCPU(params) @@ -184,7 +185,7 @@ FullO3CPU::FullO3CPU(DerivO3CPUParams *params) scoreboard(params->numThreads, TheISA::NumIntRegs, params->numPhysIntRegs, TheISA::NumFloatRegs, params->numPhysFloatRegs, - TheISA::NumMiscRegs * number_of_threads, + TheISA::NumMiscRegs * numThreads, TheISA::ZeroReg), timeBuffer(params->backComSize, params->forwardComSize), @@ -202,8 +203,7 @@ FullO3CPU::FullO3CPU(DerivO3CPUParams *params) physmem(system->physmem), #endif // FULL_SYSTEM drainCount(0), - deferRegistration(params->defer_registration), - numThreads(number_of_threads) + deferRegistration(params->defer_registration) { if (!deferRegistration) { _status = Running; @@ -224,8 +224,8 @@ FullO3CPU::FullO3CPU(DerivO3CPUParams *params) #endif // USE_CHECKER #if !FULL_SYSTEM - thread.resize(number_of_threads); - tids.resize(number_of_threads); + thread.resize(numThreads); + tids.resize(numThreads); #endif // The stages also need their CPU pointer setup. However this @@ -263,7 +263,7 @@ FullO3CPU::FullO3CPU(DerivO3CPUParams *params) rename.setCommitStage(&commit); #if !FULL_SYSTEM - int active_threads = params->workload.size(); + ThreadID active_threads = params->workload.size(); if (active_threads > Impl::MaxThreads) { panic("Workload Size too large. Increase the 'MaxThreads'" @@ -271,7 +271,7 @@ FullO3CPU::FullO3CPU(DerivO3CPUParams *params) "edit your workload size."); } #else - int active_threads = 1; + ThreadID active_threads = 1; #endif //Make Sure That this a Valid Architeture @@ -285,7 +285,7 @@ FullO3CPU::FullO3CPU(DerivO3CPUParams *params) PhysRegIndex lreg_idx = 0; PhysRegIndex freg_idx = params->numPhysIntRegs; //Index to 1 after int regs - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { bool bindRegs = (tid <= active_threads - 1); commitRenameMap[tid].init(TheISA::NumIntRegs, @@ -328,9 +328,8 @@ FullO3CPU::FullO3CPU(DerivO3CPUParams *params) commit.setRenameMap(commitRenameMap); // Give renameMap & rename stage access to the freeList; - for (int i=0; i < numThreads; i++) { - renameMap[i].setFreeList(&freeList); - } + for (ThreadID tid = 0; tid < numThreads; tid++) + renameMap[tid].setFreeList(&freeList); rename.setFreeList(&freeList); // Setup the ROB for whichever stages need it. @@ -339,11 +338,11 @@ FullO3CPU::FullO3CPU(DerivO3CPUParams *params) lastRunningCycle = curTick; lastActivatedCycle = -1; - +#if 0 // Give renameMap & rename stage access to the freeList; - //for (int i=0; i < numThreads; i++) { - //globalSeqNum[i] = 1; - //} + for (ThreadID tid = 0; tid < numThreads; tid++) + globalSeqNum[tid] = 1; +#endif contextSwitch = false; DPRINTF(O3CPU, "Creating O3CPU object.\n"); @@ -351,30 +350,30 @@ FullO3CPU::FullO3CPU(DerivO3CPUParams *params) // Setup any thread state. this->thread.resize(this->numThreads); - for (int i = 0; i < this->numThreads; ++i) { + for (ThreadID tid = 0; tid < this->numThreads; ++tid) { #if FULL_SYSTEM // SMT is not supported in FS mode yet. assert(this->numThreads == 1); - this->thread[i] = new Thread(this, 0); + this->thread[tid] = new Thread(this, 0); #else - if (i < params->workload.size()) { + if (tid < params->workload.size()) { DPRINTF(O3CPU, "Workload[%i] process is %#x", - i, this->thread[i]); - this->thread[i] = new typename FullO3CPU::Thread( + tid, this->thread[tid]); + this->thread[tid] = new typename FullO3CPU::Thread( (typename Impl::O3CPU *)(this), - i, params->workload[i], i); + tid, params->workload[tid], tid); - //usedTids[i] = true; - //threadMap[i] = i; + //usedTids[tid] = true; + //threadMap[tid] = tid; } else { //Allocate Empty thread so M5 can use later //when scheduling threads to CPU Process* dummy_proc = NULL; - this->thread[i] = new typename FullO3CPU::Thread( + this->thread[tid] = new typename FullO3CPU::Thread( (typename Impl::O3CPU *)(this), - i, dummy_proc, i); - //usedTids[i] = false; + tid, dummy_proc, tid); + //usedTids[tid] = false; } #endif // !FULL_SYSTEM @@ -396,22 +395,21 @@ FullO3CPU::FullO3CPU(DerivO3CPUParams *params) o3_tc->cpu = (typename Impl::O3CPU *)(this); assert(o3_tc->cpu); - o3_tc->thread = this->thread[i]; + o3_tc->thread = this->thread[tid]; #if FULL_SYSTEM // Setup quiesce event. - this->thread[i]->quiesceEvent = new EndQuiesceEvent(tc); + this->thread[tid]->quiesceEvent = new EndQuiesceEvent(tc); #endif // Give the thread the TC. - this->thread[i]->tc = tc; + this->thread[tid]->tc = tc; // Add the TC to the CPU's list of TC's. this->threadContexts.push_back(tc); } - for (int i=0; i < this->numThreads; i++) { - this->thread[i]->setFuncExeInst(0); - } + for (ThreadID tid = 0; tid < this->numThreads; tid++) + this->thread[tid]->setFuncExeInst(0); lockAddr = 0; lockFlag = false; @@ -565,19 +563,19 @@ FullO3CPU::init() // Set inSyscall so that the CPU doesn't squash when initially // setting up registers. - for (int i = 0; i < number_of_threads; ++i) - thread[i]->inSyscall = true; + for (ThreadID tid = 0; tid < numThreads; ++tid) + thread[tid]->inSyscall = true; #if FULL_SYSTEM - for (int tid=0; tid < number_of_threads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { ThreadContext *src_tc = threadContexts[tid]; TheISA::initCPU(src_tc, src_tc->contextId()); } #endif // Clear inSyscall. - for (int i = 0; i < number_of_threads; ++i) - thread[i]->inSyscall = false; + for (int tid = 0; tid < numThreads; ++tid) + thread[tid]->inSyscall = false; // Initialize stages. fetch.initStage(); @@ -590,9 +588,9 @@ FullO3CPU::init() template void -FullO3CPU::activateThread(unsigned tid) +FullO3CPU::activateThread(ThreadID tid) { - std::list::iterator isActive = + list::iterator isActive = std::find(activeThreads.begin(), activeThreads.end(), tid); DPRINTF(O3CPU, "[tid:%i]: Calling activate thread.\n", tid); @@ -607,10 +605,10 @@ FullO3CPU::activateThread(unsigned tid) template void -FullO3CPU::deactivateThread(unsigned tid) +FullO3CPU::deactivateThread(ThreadID tid) { //Remove From Active List, if Active - std::list::iterator thread_it = + list::iterator thread_it = std::find(activeThreads.begin(), activeThreads.end(), tid); DPRINTF(O3CPU, "[tid:%i]: Calling deactivate thread.\n", tid); @@ -622,9 +620,22 @@ FullO3CPU::deactivateThread(unsigned tid) } } +template +Counter +FullO3CPU::totalInstructions() const +{ + Counter total(0); + + ThreadID size = thread.size(); + for (ThreadID i = 0; i < size; i++) + total += thread[i]->numInst; + + return total; +} + template void -FullO3CPU::activateContext(int tid, int delay) +FullO3CPU::activateContext(ThreadID tid, int delay) { // Needs to set each stage to running as well. if (delay){ @@ -651,7 +662,7 @@ FullO3CPU::activateContext(int tid, int delay) template bool -FullO3CPU::deallocateContext(int tid, bool remove, int delay) +FullO3CPU::deallocateContext(ThreadID tid, bool remove, int delay) { // Schedule removal of thread data from CPU if (delay){ @@ -669,7 +680,7 @@ FullO3CPU::deallocateContext(int tid, bool remove, int delay) template void -FullO3CPU::suspendContext(int tid) +FullO3CPU::suspendContext(ThreadID tid) { DPRINTF(O3CPU,"[tid: %i]: Suspending Thread Context.\n", tid); bool deallocated = deallocateContext(tid, false, 1); @@ -682,7 +693,7 @@ FullO3CPU::suspendContext(int tid) template void -FullO3CPU::haltContext(int tid) +FullO3CPU::haltContext(ThreadID tid) { //For now, this is the same as deallocate DPRINTF(O3CPU,"[tid:%i]: Halt Context called. Deallocating", tid); @@ -691,7 +702,7 @@ FullO3CPU::haltContext(int tid) template void -FullO3CPU::insertThread(unsigned tid) +FullO3CPU::insertThread(ThreadID tid) { DPRINTF(O3CPU,"[tid:%i] Initializing thread into CPU"); // Will change now that the PC and thread state is internal to the CPU @@ -737,7 +748,7 @@ FullO3CPU::insertThread(unsigned tid) template void -FullO3CPU::removeThread(unsigned tid) +FullO3CPU::removeThread(ThreadID tid) { DPRINTF(O3CPU,"[tid:%i] Removing thread context from CPU.\n", tid); @@ -796,7 +807,7 @@ FullO3CPU::removeThread(unsigned tid) template void -FullO3CPU::activateWhenReady(int tid) +FullO3CPU::activateWhenReady(ThreadID tid) { DPRINTF(O3CPU,"[tid:%i]: Checking if resources are available for incoming" "(e.g. PhysRegs/ROB/IQ/LSQ) \n", @@ -855,7 +866,7 @@ FullO3CPU::activateWhenReady(int tid) #if FULL_SYSTEM template Fault -FullO3CPU::hwrei(unsigned tid) +FullO3CPU::hwrei(ThreadID tid) { #if THE_ISA == ALPHA_ISA // Need to clear the lock flag upon returning from an interrupt. @@ -870,7 +881,7 @@ FullO3CPU::hwrei(unsigned tid) template bool -FullO3CPU::simPalCheck(int palFunc, unsigned tid) +FullO3CPU::simPalCheck(int palFunc, ThreadID tid) { #if THE_ISA == ALPHA_ISA if (this->thread[tid]->kernelStats) @@ -925,14 +936,15 @@ FullO3CPU::updateMemPorts() { // Update all ThreadContext's memory ports (Functional/Virtual // Ports) - for (int i = 0; i < thread.size(); ++i) + ThreadID size = thread.size(); + for (ThreadID i = 0; i < size; ++i) thread[i]->connectMemPorts(thread[i]->getTC()); } #endif template void -FullO3CPU::trap(Fault fault, unsigned tid) +FullO3CPU::trap(Fault fault, ThreadID tid) { // Pass the thread's TC into the invoke method. fault->invoke(this->threadContexts[tid]); @@ -942,7 +954,7 @@ FullO3CPU::trap(Fault fault, unsigned tid) template void -FullO3CPU::syscall(int64_t callnum, int tid) +FullO3CPU::syscall(int64_t callnum, ThreadID tid) { DPRINTF(O3CPU, "[tid:%i] Executing syscall().\n\n", tid); @@ -977,7 +989,8 @@ FullO3CPU::serialize(std::ostream &os) // get instantiated multiple times (causes a panic in statistics). static SimpleThread temp; - for (int i = 0; i < thread.size(); i++) { + ThreadID size = thread.size(); + for (ThreadID i = 0; i < size; i++) { nameOut(os, csprintf("%s.xc.%i", name(), i)); temp.copyTC(thread[i]->getTC()); temp.serialize(os); @@ -998,7 +1011,8 @@ FullO3CPU::unserialize(Checkpoint *cp, const std::string §ion) // get instantiated multiple times (causes a panic in statistics). static SimpleThread temp; - for (int i = 0; i < thread.size(); i++) { + ThreadID size = thread.size(); + for (ThreadID i = 0; i < size; i++) { temp.copyTC(thread[i]->getTC()); temp.unserialize(cp, csprintf("%s.xc.%i", section, i)); thread[i]->getTC()->copyArchRegs(temp.getTC()); @@ -1134,9 +1148,9 @@ FullO3CPU::takeOverFrom(BaseCPU *oldCPU) // @todo: Figure out how to properly select the tid to put onto // the active threads list. - int tid = 0; + ThreadID tid = 0; - std::list::iterator isActive = + list::iterator isActive = std::find(activeThreads.begin(), activeThreads.end(), tid); if (isActive == activeThreads.end()) { @@ -1150,7 +1164,8 @@ FullO3CPU::takeOverFrom(BaseCPU *oldCPU) // Set all statuses to active, schedule the CPU's tick event. // @todo: Fix up statuses so this is handled properly - for (int i = 0; i < threadContexts.size(); ++i) { + ThreadID size = threadContexts.size(); + for (ThreadID i = 0; i < size; ++i) { ThreadContext *tc = threadContexts[i]; if (tc->status() == ThreadContext::Active && _status != Running) { _status = Running; @@ -1163,14 +1178,14 @@ FullO3CPU::takeOverFrom(BaseCPU *oldCPU) template TheISA::MiscReg -FullO3CPU::readMiscRegNoEffect(int misc_reg, unsigned tid) +FullO3CPU::readMiscRegNoEffect(int misc_reg, ThreadID tid) { return this->regFile.readMiscRegNoEffect(misc_reg, tid); } template TheISA::MiscReg -FullO3CPU::readMiscReg(int misc_reg, unsigned tid) +FullO3CPU::readMiscReg(int misc_reg, ThreadID tid) { return this->regFile.readMiscReg(misc_reg, tid); } @@ -1178,7 +1193,7 @@ FullO3CPU::readMiscReg(int misc_reg, unsigned tid) template void FullO3CPU::setMiscRegNoEffect(int misc_reg, - const TheISA::MiscReg &val, unsigned tid) + const TheISA::MiscReg &val, ThreadID tid) { this->regFile.setMiscRegNoEffect(misc_reg, val, tid); } @@ -1186,7 +1201,7 @@ FullO3CPU::setMiscRegNoEffect(int misc_reg, template void FullO3CPU::setMiscReg(int misc_reg, - const TheISA::MiscReg &val, unsigned tid) + const TheISA::MiscReg &val, ThreadID tid) { this->regFile.setMiscReg(misc_reg, val, tid); } @@ -1263,7 +1278,7 @@ FullO3CPU::setFloatRegBits(int reg_idx, FloatRegBits val) template uint64_t -FullO3CPU::readArchIntReg(int reg_idx, unsigned tid) +FullO3CPU::readArchIntReg(int reg_idx, ThreadID tid) { PhysRegIndex phys_reg = commitRenameMap[tid].lookup(reg_idx); @@ -1272,7 +1287,7 @@ FullO3CPU::readArchIntReg(int reg_idx, unsigned tid) template float -FullO3CPU::readArchFloatRegSingle(int reg_idx, unsigned tid) +FullO3CPU::readArchFloatRegSingle(int reg_idx, ThreadID tid) { int idx = reg_idx + TheISA::NumIntRegs; PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx); @@ -1282,7 +1297,7 @@ FullO3CPU::readArchFloatRegSingle(int reg_idx, unsigned tid) template double -FullO3CPU::readArchFloatRegDouble(int reg_idx, unsigned tid) +FullO3CPU::readArchFloatRegDouble(int reg_idx, ThreadID tid) { int idx = reg_idx + TheISA::NumIntRegs; PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx); @@ -1292,7 +1307,7 @@ FullO3CPU::readArchFloatRegDouble(int reg_idx, unsigned tid) template uint64_t -FullO3CPU::readArchFloatRegInt(int reg_idx, unsigned tid) +FullO3CPU::readArchFloatRegInt(int reg_idx, ThreadID tid) { int idx = reg_idx + TheISA::NumIntRegs; PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx); @@ -1302,7 +1317,7 @@ FullO3CPU::readArchFloatRegInt(int reg_idx, unsigned tid) template void -FullO3CPU::setArchIntReg(int reg_idx, uint64_t val, unsigned tid) +FullO3CPU::setArchIntReg(int reg_idx, uint64_t val, ThreadID tid) { PhysRegIndex phys_reg = commitRenameMap[tid].lookup(reg_idx); @@ -1311,7 +1326,7 @@ FullO3CPU::setArchIntReg(int reg_idx, uint64_t val, unsigned tid) template void -FullO3CPU::setArchFloatRegSingle(int reg_idx, float val, unsigned tid) +FullO3CPU::setArchFloatRegSingle(int reg_idx, float val, ThreadID tid) { int idx = reg_idx + TheISA::NumIntRegs; PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx); @@ -1321,7 +1336,7 @@ FullO3CPU::setArchFloatRegSingle(int reg_idx, float val, unsigned tid) template void -FullO3CPU::setArchFloatRegDouble(int reg_idx, double val, unsigned tid) +FullO3CPU::setArchFloatRegDouble(int reg_idx, double val, ThreadID tid) { int idx = reg_idx + TheISA::NumIntRegs; PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx); @@ -1331,7 +1346,7 @@ FullO3CPU::setArchFloatRegDouble(int reg_idx, double val, unsigned tid) template void -FullO3CPU::setArchFloatRegInt(int reg_idx, uint64_t val, unsigned tid) +FullO3CPU::setArchFloatRegInt(int reg_idx, uint64_t val, ThreadID tid) { int idx = reg_idx + TheISA::NumIntRegs; PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx); @@ -1341,77 +1356,77 @@ FullO3CPU::setArchFloatRegInt(int reg_idx, uint64_t val, unsigned tid) template uint64_t -FullO3CPU::readPC(unsigned tid) +FullO3CPU::readPC(ThreadID tid) { return commit.readPC(tid); } template void -FullO3CPU::setPC(Addr new_PC,unsigned tid) +FullO3CPU::setPC(Addr new_PC, ThreadID tid) { commit.setPC(new_PC, tid); } template uint64_t -FullO3CPU::readMicroPC(unsigned tid) +FullO3CPU::readMicroPC(ThreadID tid) { return commit.readMicroPC(tid); } template void -FullO3CPU::setMicroPC(Addr new_PC,unsigned tid) +FullO3CPU::setMicroPC(Addr new_PC, ThreadID tid) { commit.setMicroPC(new_PC, tid); } template uint64_t -FullO3CPU::readNextPC(unsigned tid) +FullO3CPU::readNextPC(ThreadID tid) { return commit.readNextPC(tid); } template void -FullO3CPU::setNextPC(uint64_t val,unsigned tid) +FullO3CPU::setNextPC(uint64_t val, ThreadID tid) { commit.setNextPC(val, tid); } template uint64_t -FullO3CPU::readNextNPC(unsigned tid) +FullO3CPU::readNextNPC(ThreadID tid) { return commit.readNextNPC(tid); } template void -FullO3CPU::setNextNPC(uint64_t val,unsigned tid) +FullO3CPU::setNextNPC(uint64_t val, ThreadID tid) { commit.setNextNPC(val, tid); } template uint64_t -FullO3CPU::readNextMicroPC(unsigned tid) +FullO3CPU::readNextMicroPC(ThreadID tid) { return commit.readNextMicroPC(tid); } template void -FullO3CPU::setNextMicroPC(Addr new_PC,unsigned tid) +FullO3CPU::setNextMicroPC(Addr new_PC, ThreadID tid) { commit.setNextMicroPC(new_PC, tid); } template void -FullO3CPU::squashFromTC(unsigned tid) +FullO3CPU::squashFromTC(ThreadID tid) { this->thread[tid]->inSyscall = true; this->commit.generateTCEvent(tid); @@ -1428,7 +1443,7 @@ FullO3CPU::addInst(DynInstPtr &inst) template void -FullO3CPU::instDone(unsigned tid) +FullO3CPU::instDone(ThreadID tid) { // Keep an instruction count. thread[tid]->numInst++; @@ -1465,7 +1480,7 @@ FullO3CPU::removeFrontInst(DynInstPtr &inst) template void -FullO3CPU::removeInstsNotInROB(unsigned tid) +FullO3CPU::removeInstsNotInROB(ThreadID tid) { DPRINTF(O3CPU, "Thread %i: Deleting instructions from instruction" " list.\n", tid); @@ -1510,8 +1525,7 @@ FullO3CPU::removeInstsNotInROB(unsigned tid) template void -FullO3CPU::removeInstsUntil(const InstSeqNum &seq_num, - unsigned tid) +FullO3CPU::removeInstsUntil(const InstSeqNum &seq_num, ThreadID tid) { assert(!instList.empty()); @@ -1540,7 +1554,7 @@ FullO3CPU::removeInstsUntil(const InstSeqNum &seq_num, template inline void -FullO3CPU::squashInstIt(const ListIt &instIt, const unsigned &tid) +FullO3CPU::squashInstIt(const ListIt &instIt, ThreadID tid) { if ((*instIt)->threadNumber == tid) { DPRINTF(O3CPU, "Squashing instruction, " @@ -1646,17 +1660,17 @@ FullO3CPU::wakeup() #endif template -int +ThreadID FullO3CPU::getFreeTid() { - for (int i=0; i < numThreads; i++) { - if (!tids[i]) { - tids[i] = true; - return i; + for (ThreadID tid = 0; tid < numThreads; tid++) { + if (!tids[tid]) { + tids[tid] = true; + return tid; } } - return -1; + return InvalidThreadID; } template @@ -1667,7 +1681,8 @@ FullO3CPU::doContextSwitch() //ADD CODE TO DEACTIVE THREAD HERE (???) - for (int tid=0; tid < cpuWaitList.size(); tid++) { + ThreadID size = cpuWaitList.size(); + for (ThreadID tid = 0; tid < size; tid++) { activateWhenReady(tid); } @@ -1680,12 +1695,11 @@ template void FullO3CPU::updateThreadPriority() { - if (activeThreads.size() > 1) - { + if (activeThreads.size() > 1) { //DEFAULT TO ROUND ROBIN SCHEME //e.g. Move highest priority to end of thread list - std::list::iterator list_begin = activeThreads.begin(); - std::list::iterator list_end = activeThreads.end(); + list::iterator list_begin = activeThreads.begin(); + list::iterator list_end = activeThreads.end(); unsigned high_thread = *list_begin; diff --git a/src/cpu/o3/cpu.hh b/src/cpu/o3/cpu.hh index 9eead4f49..5cf27df75 100644 --- a/src/cpu/o3/cpu.hh +++ b/src/cpu/o3/cpu.hh @@ -155,7 +155,7 @@ class FullO3CPU : public BaseO3CPU { private: /** Number of Thread to Activate */ - int tid; + ThreadID tid; /** Pointer to the CPU. */ FullO3CPU *cpu; @@ -175,7 +175,8 @@ class FullO3CPU : public BaseO3CPU }; /** Schedule thread to activate , regardless of its current state. */ - void scheduleActivateThreadEvent(int tid, int delay) + void + scheduleActivateThreadEvent(ThreadID tid, int delay) { // Schedule thread to activate, regardless of its current state. if (activateThreadEvent[tid].squashed()) @@ -187,7 +188,8 @@ class FullO3CPU : public BaseO3CPU } /** Unschedule actiavte thread event, regardless of its current state. */ - void unscheduleActivateThreadEvent(int tid) + void + unscheduleActivateThreadEvent(ThreadID tid) { if (activateThreadEvent[tid].scheduled()) activateThreadEvent[tid].squash(); @@ -200,7 +202,7 @@ class FullO3CPU : public BaseO3CPU { private: /** Number of Thread to deactivate */ - int tid; + ThreadID tid; /** Should the thread be removed from the CPU? */ bool remove; @@ -226,7 +228,8 @@ class FullO3CPU : public BaseO3CPU }; /** Schedule cpu to deallocate thread context.*/ - void scheduleDeallocateContextEvent(int tid, bool remove, int delay) + void + scheduleDeallocateContextEvent(ThreadID tid, bool remove, int delay) { // Schedule thread to activate, regardless of its current state. if (deallocateContextEvent[tid].squashed()) @@ -238,7 +241,8 @@ class FullO3CPU : public BaseO3CPU } /** Unschedule thread deallocation in CPU */ - void unscheduleDeallocateContextEvent(int tid) + void + unscheduleDeallocateContextEvent(ThreadID tid) { if (deallocateContextEvent[tid].scheduled()) deallocateContextEvent[tid].squash(); @@ -288,46 +292,38 @@ class FullO3CPU : public BaseO3CPU { return activeThreads.size(); } /** Add Thread to Active Threads List */ - void activateThread(unsigned tid); + void activateThread(ThreadID tid); /** Remove Thread from Active Threads List */ - void deactivateThread(unsigned tid); + void deactivateThread(ThreadID tid); /** Setup CPU to insert a thread's context */ - void insertThread(unsigned tid); + void insertThread(ThreadID tid); /** Remove all of a thread's context from CPU */ - void removeThread(unsigned tid); + void removeThread(ThreadID tid); /** Count the Total Instructions Committed in the CPU. */ - virtual Counter totalInstructions() const - { - Counter total(0); - - for (int i=0; i < thread.size(); i++) - total += thread[i]->numInst; - - return total; - } + virtual Counter totalInstructions() const; /** Add Thread to Active Threads List. */ - void activateContext(int tid, int delay); + void activateContext(ThreadID tid, int delay); /** Remove Thread from Active Threads List */ - void suspendContext(int tid); + void suspendContext(ThreadID tid); /** Remove Thread from Active Threads List && * Possibly Remove Thread Context from CPU. */ - bool deallocateContext(int tid, bool remove, int delay = 1); + bool deallocateContext(ThreadID tid, bool remove, int delay = 1); /** Remove Thread from Active Threads List && * Remove Thread Context from CPU. */ - void haltContext(int tid); + void haltContext(ThreadID tid); /** Activate a Thread When CPU Resources are Available. */ - void activateWhenReady(int tid); + void activateWhenReady(ThreadID tid); /** Add or Remove a Thread Context in the CPU. */ void doContextSwitch(); @@ -346,7 +342,7 @@ class FullO3CPU : public BaseO3CPU /** Executes a syscall. * @todo: Determine if this needs to be virtual. */ - void syscall(int64_t callnum, int tid); + void syscall(int64_t callnum, ThreadID tid); #endif /** Starts draining the CPU's pipeline of all instructions in @@ -370,13 +366,13 @@ class FullO3CPU : public BaseO3CPU { return globalSeqNum++; } /** Traps to handle given fault. */ - void trap(Fault fault, unsigned tid); + void trap(Fault fault, ThreadID tid); #if FULL_SYSTEM /** HW return from error interrupt. */ - Fault hwrei(unsigned tid); + Fault hwrei(ThreadID tid); - bool simPalCheck(int palFunc, unsigned tid); + bool simPalCheck(int palFunc, ThreadID tid); /** Returns the Fault for any valid interrupt. */ Fault getInterrupts(); @@ -398,19 +394,19 @@ class FullO3CPU : public BaseO3CPU bool validDataAddr(Addr addr) { return true; } /** Get instruction asid. */ - int getInstAsid(unsigned tid) + int getInstAsid(ThreadID tid) { return regFile.miscRegs[tid].getInstAsid(); } /** Get data asid. */ - int getDataAsid(unsigned tid) + int getDataAsid(ThreadID tid) { return regFile.miscRegs[tid].getDataAsid(); } #else /** Get instruction asid. */ - int getInstAsid(unsigned tid) + int getInstAsid(ThreadID tid) { return thread[tid]->getInstAsid(); } /** Get data asid. */ - int getDataAsid(unsigned tid) + int getDataAsid(ThreadID tid) { return thread[tid]->getDataAsid(); } #endif @@ -418,21 +414,22 @@ class FullO3CPU : public BaseO3CPU /** Register accessors. Index refers to the physical register index. */ /** Reads a miscellaneous register. */ - TheISA::MiscReg readMiscRegNoEffect(int misc_reg, unsigned tid); + TheISA::MiscReg readMiscRegNoEffect(int misc_reg, ThreadID tid); /** Reads a misc. register, including any side effects the read * might have as defined by the architecture. */ - TheISA::MiscReg readMiscReg(int misc_reg, unsigned tid); + TheISA::MiscReg readMiscReg(int misc_reg, ThreadID tid); /** Sets a miscellaneous register. */ - void setMiscRegNoEffect(int misc_reg, const TheISA::MiscReg &val, unsigned tid); + void setMiscRegNoEffect(int misc_reg, const TheISA::MiscReg &val, + ThreadID tid); /** Sets a misc. register, including any side effects the write * might have as defined by the architecture. */ void setMiscReg(int misc_reg, const TheISA::MiscReg &val, - unsigned tid); + ThreadID tid); uint64_t readIntReg(int reg_idx); @@ -454,62 +451,62 @@ class FullO3CPU : public BaseO3CPU void setFloatRegBits(int reg_idx, TheISA::FloatRegBits val, int width); - uint64_t readArchIntReg(int reg_idx, unsigned tid); + uint64_t readArchIntReg(int reg_idx, ThreadID tid); - float readArchFloatRegSingle(int reg_idx, unsigned tid); + float readArchFloatRegSingle(int reg_idx, ThreadID tid); - double readArchFloatRegDouble(int reg_idx, unsigned tid); + double readArchFloatRegDouble(int reg_idx, ThreadID tid); - uint64_t readArchFloatRegInt(int reg_idx, unsigned tid); + uint64_t readArchFloatRegInt(int reg_idx, ThreadID tid); /** Architectural register accessors. Looks up in the commit * rename table to obtain the true physical index of the * architected register first, then accesses that physical * register. */ - void setArchIntReg(int reg_idx, uint64_t val, unsigned tid); + void setArchIntReg(int reg_idx, uint64_t val, ThreadID tid); - void setArchFloatRegSingle(int reg_idx, float val, unsigned tid); + void setArchFloatRegSingle(int reg_idx, float val, ThreadID tid); - void setArchFloatRegDouble(int reg_idx, double val, unsigned tid); + void setArchFloatRegDouble(int reg_idx, double val, ThreadID tid); - void setArchFloatRegInt(int reg_idx, uint64_t val, unsigned tid); + void setArchFloatRegInt(int reg_idx, uint64_t val, ThreadID tid); /** Reads the commit PC of a specific thread. */ - Addr readPC(unsigned tid); + Addr readPC(ThreadID tid); /** Sets the commit PC of a specific thread. */ - void setPC(Addr new_PC, unsigned tid); + void setPC(Addr new_PC, ThreadID tid); /** Reads the commit micro PC of a specific thread. */ - Addr readMicroPC(unsigned tid); + Addr readMicroPC(ThreadID tid); /** Sets the commmit micro PC of a specific thread. */ - void setMicroPC(Addr new_microPC, unsigned tid); + void setMicroPC(Addr new_microPC, ThreadID tid); /** Reads the next PC of a specific thread. */ - Addr readNextPC(unsigned tid); + Addr readNextPC(ThreadID tid); /** Sets the next PC of a specific thread. */ - void setNextPC(Addr val, unsigned tid); + void setNextPC(Addr val, ThreadID tid); /** Reads the next NPC of a specific thread. */ - Addr readNextNPC(unsigned tid); + Addr readNextNPC(ThreadID tid); /** Sets the next NPC of a specific thread. */ - void setNextNPC(Addr val, unsigned tid); + void setNextNPC(Addr val, ThreadID tid); /** Reads the commit next micro PC of a specific thread. */ - Addr readNextMicroPC(unsigned tid); + Addr readNextMicroPC(ThreadID tid); /** Sets the commit next micro PC of a specific thread. */ - void setNextMicroPC(Addr val, unsigned tid); + void setNextMicroPC(Addr val, ThreadID tid); /** Initiates a squash of all in-flight instructions for a given * thread. The source of the squash is an external update of * state through the TC. */ - void squashFromTC(unsigned tid); + void squashFromTC(ThreadID tid); /** Function to add instruction onto the head of the list of the * instructions. Used when new instructions are fetched. @@ -517,7 +514,7 @@ class FullO3CPU : public BaseO3CPU ListIt addInst(DynInstPtr &inst); /** Function to tell the CPU that an instruction has completed. */ - void instDone(unsigned tid); + void instDone(ThreadID tid); /** Add Instructions to the CPU Remove List*/ void addToRemoveList(DynInstPtr &inst); @@ -529,13 +526,13 @@ class FullO3CPU : public BaseO3CPU /** Remove all instructions that are not currently in the ROB. * There's also an option to not squash delay slot instructions.*/ - void removeInstsNotInROB(unsigned tid); + void removeInstsNotInROB(ThreadID tid); /** Remove all instructions younger than the given sequence number. */ - void removeInstsUntil(const InstSeqNum &seq_num,unsigned tid); + void removeInstsUntil(const InstSeqNum &seq_num, ThreadID tid); /** Removes the instruction pointed to by the iterator. */ - inline void squashInstIt(const ListIt &instIt, const unsigned &tid); + inline void squashInstIt(const ListIt &instIt, ThreadID tid); /** Cleans up all instructions on the remove list. */ void cleanUpRemovedInsts(); @@ -601,7 +598,7 @@ class FullO3CPU : public BaseO3CPU typename CPUPolicy::ROB rob; /** Active Threads List */ - std::list activeThreads; + std::list activeThreads; /** Integer Register Scoreboard */ Scoreboard scoreboard; @@ -674,11 +671,12 @@ class FullO3CPU : public BaseO3CPU #endif /** Gets a free thread id. Use if thread ids change across system. */ - int getFreeTid(); + ThreadID getFreeTid(); public: /** Returns a pointer to a thread context. */ - ThreadContext *tcBase(unsigned tid) + ThreadContext * + tcBase(ThreadID tid) { return thread[tid]->getTC(); } @@ -726,14 +724,11 @@ class FullO3CPU : public BaseO3CPU /** The cycle that the CPU was last activated by a new thread*/ Tick lastActivatedCycle; - /** Number of Threads CPU can process */ - unsigned numThreads; - /** Mapping for system thread id to cpu id */ - std::map threadMap; + std::map threadMap; /** Available thread ids in the cpu*/ - std::vector tids; + std::vector tids; /** CPU read function, forwards read to LSQ. */ template diff --git a/src/cpu/o3/cpu_builder.cc b/src/cpu/o3/cpu_builder.cc index 77d7091ad..a433235a0 100644 --- a/src/cpu/o3/cpu_builder.cc +++ b/src/cpu/o3/cpu_builder.cc @@ -49,11 +49,11 @@ DerivO3CPUParams::create() { #if FULL_SYSTEM // Full-system only supports a single thread for the moment. - int actual_num_threads = 1; + ThreadID actual_num_threads = 1; #else // In non-full-system mode, we infer the number of threads from // the workload if it's not explicitly specified. - int actual_num_threads = + ThreadID actual_num_threads = (numThreads >= workload.size()) ? numThreads : workload.size(); if (workload.size() == 0) { diff --git a/src/cpu/o3/decode.hh b/src/cpu/o3/decode.hh index 294b5b623..e272b661f 100644 --- a/src/cpu/o3/decode.hh +++ b/src/cpu/o3/decode.hh @@ -105,7 +105,7 @@ class DefaultDecode void setFetchQueue(TimeBuffer *fq_ptr); /** Sets pointer to list of active threads. */ - void setActiveThreads(std::list *at_ptr); + void setActiveThreads(std::list *at_ptr); /** Drains the decode stage. */ bool drain(); @@ -129,20 +129,20 @@ class DefaultDecode * change (ie switching from from blocking to unblocking). * @param tid Thread id to decode instructions from. */ - void decode(bool &status_change, unsigned tid); + void decode(bool &status_change, ThreadID tid); /** Processes instructions from fetch and passes them on to rename. * Decoding of instructions actually happens when they are created in * fetch, so this function mostly checks if PC-relative branches are * correct. */ - void decodeInsts(unsigned tid); + void decodeInsts(ThreadID tid); private: /** Inserts a thread's instructions into the skid buffer, to be decoded * once decode unblocks. */ - void skidInsert(unsigned tid); + void skidInsert(ThreadID tid); /** Returns if all of the skid buffers are empty. */ bool skidsEmpty(); @@ -156,13 +156,13 @@ class DefaultDecode void sortInsts(); /** Reads all stall signals from the backwards communication timebuffer. */ - void readStallSignals(unsigned tid); + void readStallSignals(ThreadID tid); /** Checks all input signals and updates decode's status appropriately. */ - bool checkSignalsAndUpdate(unsigned tid); + bool checkSignalsAndUpdate(ThreadID tid); /** Checks all stall signals, and returns if any are true. */ - bool checkStall(unsigned tid) const; + bool checkStall(ThreadID tid) const; /** Returns if there any instructions from fetch on this cycle. */ inline bool fetchInstsValid(); @@ -171,24 +171,24 @@ class DefaultDecode * become blocked. * @return Returns true if there is a status change. */ - bool block(unsigned tid); + bool block(ThreadID tid); /** Switches decode to unblocking if the skid buffer is empty, and * signals back that decode has unblocked. * @return Returns true if there is a status change. */ - bool unblock(unsigned tid); + bool unblock(ThreadID tid); /** Squashes if there is a PC-relative branch that was predicted * incorrectly. Sends squash information back to fetch. */ - void squash(DynInstPtr &inst, unsigned tid); + void squash(DynInstPtr &inst, ThreadID tid); public: /** Squashes due to commit signalling a squash. Changes status to * squashing and clears block/unblock signals as needed. */ - unsigned squash(unsigned tid); + unsigned squash(ThreadID tid); private: // Interfaces to objects outside of decode. @@ -263,10 +263,10 @@ class DefaultDecode unsigned toRenameIndex; /** number of Active Threads*/ - unsigned numThreads; + ThreadID numThreads; /** List of active thread ids */ - std::list *activeThreads; + std::list *activeThreads; /** Number of branches in flight. */ unsigned branchCount[Impl::MaxThreads]; diff --git a/src/cpu/o3/decode_impl.hh b/src/cpu/o3/decode_impl.hh index d92e97f61..86f87991c 100644 --- a/src/cpu/o3/decode_impl.hh +++ b/src/cpu/o3/decode_impl.hh @@ -29,9 +29,10 @@ */ #include "cpu/o3/decode.hh" - #include "params/DerivO3CPU.hh" +using namespace std; + template DefaultDecode::DefaultDecode(O3CPU *_cpu, DerivO3CPUParams *params) : cpu(_cpu), @@ -45,12 +46,12 @@ DefaultDecode::DefaultDecode(O3CPU *_cpu, DerivO3CPUParams *params) _status = Inactive; // Setup status, make sure stall signals are clear. - for (int i = 0; i < numThreads; ++i) { - decodeStatus[i] = Idle; + for (ThreadID tid = 0; tid < numThreads; ++tid) { + decodeStatus[tid] = Idle; - stalls[i].rename = false; - stalls[i].iew = false; - stalls[i].commit = false; + stalls[tid].rename = false; + stalls[tid].iew = false; + stalls[tid].commit = false; } // @todo: Make into a parameter @@ -148,7 +149,7 @@ DefaultDecode::setFetchQueue(TimeBuffer *fq_ptr) template void -DefaultDecode::setActiveThreads(std::list *at_ptr) +DefaultDecode::setActiveThreads(std::list *at_ptr) { activeThreads = at_ptr; } @@ -169,24 +170,24 @@ DefaultDecode::takeOverFrom() _status = Inactive; // Be sure to reset state and clear out any old instructions. - for (int i = 0; i < numThreads; ++i) { - decodeStatus[i] = Idle; + for (ThreadID tid = 0; tid < numThreads; ++tid) { + decodeStatus[tid] = Idle; - stalls[i].rename = false; - stalls[i].iew = false; - stalls[i].commit = false; - while (!insts[i].empty()) - insts[i].pop(); - while (!skidBuffer[i].empty()) - skidBuffer[i].pop(); - branchCount[i] = 0; + stalls[tid].rename = false; + stalls[tid].iew = false; + stalls[tid].commit = false; + while (!insts[tid].empty()) + insts[tid].pop(); + while (!skidBuffer[tid].empty()) + skidBuffer[tid].pop(); + branchCount[tid] = 0; } wroteToTimeBuffer = false; } template bool -DefaultDecode::checkStall(unsigned tid) const +DefaultDecode::checkStall(ThreadID tid) const { bool ret_val = false; @@ -213,7 +214,7 @@ DefaultDecode::fetchInstsValid() template bool -DefaultDecode::block(unsigned tid) +DefaultDecode::block(ThreadID tid) { DPRINTF(Decode, "[tid:%u]: Blocking.\n", tid); @@ -241,7 +242,7 @@ DefaultDecode::block(unsigned tid) template bool -DefaultDecode::unblock(unsigned tid) +DefaultDecode::unblock(ThreadID tid) { // Decode is done unblocking only if the skid buffer is empty. if (skidBuffer[tid].empty()) { @@ -260,7 +261,7 @@ DefaultDecode::unblock(unsigned tid) template void -DefaultDecode::squash(DynInstPtr &inst, unsigned tid) +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); @@ -320,7 +321,7 @@ DefaultDecode::squash(DynInstPtr &inst, unsigned tid) template unsigned -DefaultDecode::squash(unsigned tid) +DefaultDecode::squash(ThreadID tid) { DPRINTF(Decode, "[tid:%i]: Squashing.\n",tid); @@ -369,7 +370,7 @@ DefaultDecode::squash(unsigned tid) template void -DefaultDecode::skidInsert(unsigned tid) +DefaultDecode::skidInsert(ThreadID tid) { DynInstPtr inst = NULL; @@ -395,11 +396,11 @@ template bool DefaultDecode::skidsEmpty() { - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!skidBuffer[tid].empty()) return false; } @@ -413,11 +414,11 @@ DefaultDecode::updateStatus() { bool any_unblocking = false; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (decodeStatus[tid] == Unblocking) { any_unblocking = true; @@ -452,8 +453,8 @@ DefaultDecode::sortInsts() { int insts_from_fetch = fromFetch->size; #ifdef DEBUG - for (int i=0; i < numThreads; i++) - assert(insts[i].empty()); + for (ThreadID tid = 0; tid < numThreads; tid++) + assert(insts[tid].empty()); #endif for (int i = 0; i < insts_from_fetch; ++i) { insts[fromFetch->insts[i]->threadNumber].push(fromFetch->insts[i]); @@ -462,7 +463,7 @@ DefaultDecode::sortInsts() template void -DefaultDecode::readStallSignals(unsigned tid) +DefaultDecode::readStallSignals(ThreadID tid) { if (fromRename->renameBlock[tid]) { stalls[tid].rename = true; @@ -494,7 +495,7 @@ DefaultDecode::readStallSignals(unsigned tid) template bool -DefaultDecode::checkSignalsAndUpdate(unsigned tid) +DefaultDecode::checkSignalsAndUpdate(ThreadID tid) { // Check if there's a squash signal, squash if there is. // Check stall signals, block if necessary. @@ -569,14 +570,14 @@ DefaultDecode::tick() toRenameIndex = 0; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); sortInsts(); //Check stall and squash signals. while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; DPRINTF(Decode,"Processing [tid:%i]\n",tid); status_change = checkSignalsAndUpdate(tid) || status_change; @@ -597,7 +598,7 @@ DefaultDecode::tick() template void -DefaultDecode::decode(bool &status_change, unsigned tid) +DefaultDecode::decode(bool &status_change, ThreadID tid) { // If status is Running or idle, // call decodeInsts() @@ -642,7 +643,7 @@ DefaultDecode::decode(bool &status_change, unsigned tid) template void -DefaultDecode::decodeInsts(unsigned tid) +DefaultDecode::decodeInsts(ThreadID tid) { // Instructions can come either from the skid buffer or the list of // instructions coming from fetch, depending on decode's status. diff --git a/src/cpu/o3/fetch.hh b/src/cpu/o3/fetch.hh index 08ccb094b..9cbc50899 100644 --- a/src/cpu/o3/fetch.hh +++ b/src/cpu/o3/fetch.hh @@ -157,7 +157,7 @@ class DefaultFetch FetchPriority fetchPolicy; /** List that has the threads organized by priority. */ - std::list priorityList; + std::list priorityList; public: /** DefaultFetch constructor. */ @@ -176,7 +176,7 @@ class DefaultFetch void setTimeBuffer(TimeBuffer *time_buffer); /** Sets pointer to list of active threads. */ - void setActiveThreads(std::list *at_ptr); + void setActiveThreads(std::list *at_ptr); /** Sets pointer to time buffer used to communicate to the next stage. */ void setFetchQueue(TimeBuffer *fq_ptr); @@ -240,21 +240,21 @@ class DefaultFetch * @param tid Thread id. * @return Any fault that occured. */ - bool fetchCacheLine(Addr fetch_PC, Fault &ret_fault, unsigned tid); + 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, unsigned tid); + const Addr &new_MicroPC, 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, - const InstSeqNum &seq_num, unsigned tid); + const InstSeqNum &seq_num, ThreadID tid); /** Checks if a thread is stalled. */ - bool checkStall(unsigned tid) const; + bool checkStall(ThreadID tid) const; /** Updates overall fetch stage status; to be called at the end of each * cycle. */ @@ -267,7 +267,7 @@ class DefaultFetch */ void squash(const Addr &new_PC, const Addr &new_NPC, const Addr &new_MicroPC, - const InstSeqNum &seq_num, unsigned tid); + const InstSeqNum &seq_num, ThreadID tid); /** Ticks the fetch stage, processing all inputs signals and fetching * as many instructions as possible. @@ -277,7 +277,7 @@ class DefaultFetch /** Checks all input signals and updates the status as necessary. * @return: Returns if the status has changed due to input signals. */ - bool checkSignalsAndUpdate(unsigned tid); + bool checkSignalsAndUpdate(ThreadID tid); /** Does the actual fetching of instructions and passing them on to the * next stage. @@ -298,19 +298,20 @@ class DefaultFetch void recvRetry(); /** Returns the appropriate thread to fetch, given the fetch policy. */ - int getFetchingThread(FetchPriority &fetch_priority); + ThreadID getFetchingThread(FetchPriority &fetch_priority); /** Returns the appropriate thread to fetch using a round robin policy. */ - int roundRobin(); + ThreadID roundRobin(); /** Returns the appropriate thread to fetch using the IQ count policy. */ - int iqCount(); + ThreadID iqCount(); /** Returns the appropriate thread to fetch using the LSQ count policy. */ - int lsqCount(); + ThreadID lsqCount(); - /** Returns the appropriate thread to fetch using the branch count policy. */ - int branchCount(); + /** Returns the appropriate thread to fetch using the branch count + * policy. */ + ThreadID branchCount(); private: /** Pointer to the O3CPU. */ @@ -400,7 +401,7 @@ class DefaultFetch PacketPtr retryPkt; /** The thread that is waiting on the cache to tell fetch to retry. */ - int retryTid; + ThreadID retryTid; /** Cache block size. */ int cacheBlkSize; @@ -424,16 +425,16 @@ class DefaultFetch Counter lastIcacheStall[Impl::MaxThreads]; /** List of Active Threads */ - std::list *activeThreads; + std::list *activeThreads; /** Number of threads. */ - unsigned numThreads; + ThreadID numThreads; /** Number of threads that are actively fetching. */ - unsigned numFetchingThreads; + ThreadID numFetchingThreads; /** Thread ID being fetched. */ - int threadFetched; + ThreadID threadFetched; /** Checks if there is an interrupt pending. If there is, fetch * must stop once it is not fetching PAL instructions. diff --git a/src/cpu/o3/fetch_impl.hh b/src/cpu/o3/fetch_impl.hh index 4d8033a8c..a76e07576 100644 --- a/src/cpu/o3/fetch_impl.hh +++ b/src/cpu/o3/fetch_impl.hh @@ -51,6 +51,8 @@ #include "sim/system.hh" #endif // FULL_SYSTEM +using namespace std; + template void DefaultFetch::IcachePort::setPeer(Port *port) @@ -122,7 +124,7 @@ DefaultFetch::DefaultFetch(O3CPU *_cpu, DerivO3CPUParams *params) fetchWidth(params->fetchWidth), cacheBlocked(false), retryPkt(NULL), - retryTid(-1), + retryTid(InvalidThreadID), numThreads(params->numThreads), numFetchingThreads(params->smtNumFetchingThreads), interruptPending(false), @@ -292,7 +294,7 @@ DefaultFetch::setTimeBuffer(TimeBuffer *time_buffer) template void -DefaultFetch::setActiveThreads(std::list *at_ptr) +DefaultFetch::setActiveThreads(std::list *at_ptr) { activeThreads = at_ptr; } @@ -312,13 +314,13 @@ void DefaultFetch::initStage() { // Setup PC and nextPC with initial state. - for (int tid = 0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { PC[tid] = cpu->readPC(tid); nextPC[tid] = cpu->readNextPC(tid); microPC[tid] = cpu->readMicroPC(tid); } - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { fetchStatus[tid] = Running; @@ -350,7 +352,7 @@ DefaultFetch::setIcache() // Create mask to get rid of offset bits. cacheBlkMask = (cacheBlkSize - 1); - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { // Create space to store a cache line. cacheData[tid] = new uint8_t[cacheBlkSize]; cacheDataPC[tid] = 0; @@ -362,7 +364,7 @@ template void DefaultFetch::processCacheCompletion(PacketPtr pkt) { - unsigned tid = pkt->req->threadId(); + ThreadID tid = pkt->req->threadId(); DPRINTF(Fetch, "[tid:%u] Waking up from cache miss.\n",tid); @@ -437,7 +439,7 @@ void DefaultFetch::takeOverFrom() { // Reset all state - for (int i = 0; i < Impl::MaxThreads; ++i) { + for (ThreadID i = 0; i < Impl::MaxThreads; ++i) { stalls[i].decode = 0; stalls[i].rename = 0; stalls[i].iew = 0; @@ -518,7 +520,7 @@ DefaultFetch::lookupAndUpdateNextPC(DynInstPtr &inst, Addr &next_PC, //would reset the micro pc to 0. next_MicroPC = 0; - int tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; Addr pred_PC = next_PC; predict_taken = branchPred.predict(inst, pred_PC, tid); @@ -560,7 +562,7 @@ DefaultFetch::lookupAndUpdateNextPC(DynInstPtr &inst, Addr &next_PC, template bool -DefaultFetch::fetchCacheLine(Addr fetch_PC, Fault &ret_fault, unsigned tid) +DefaultFetch::fetchCacheLine(Addr fetch_PC, Fault &ret_fault, ThreadID tid) { Fault fault = NoFault; @@ -637,7 +639,7 @@ DefaultFetch::fetchCacheLine(Addr fetch_PC, Fault &ret_fault, unsigned tid // exists within the cache. if (!icachePort->sendTiming(data_pkt)) { assert(retryPkt == NULL); - assert(retryTid == -1); + assert(retryTid == InvalidThreadID); DPRINTF(Fetch, "[tid:%i] Out of MSHRs!\n", tid); fetchStatus[tid] = IcacheWaitRetry; retryPkt = data_pkt; @@ -666,7 +668,7 @@ DefaultFetch::fetchCacheLine(Addr fetch_PC, Fault &ret_fault, unsigned tid template inline void DefaultFetch::doSquash(const Addr &new_PC, - const Addr &new_NPC, const Addr &new_microPC, unsigned tid) + const Addr &new_NPC, const Addr &new_microPC, ThreadID tid) { DPRINTF(Fetch, "[tid:%i]: Squashing, setting PC to: %#x, NPC to: %#x.\n", tid, new_PC, new_NPC); @@ -690,7 +692,7 @@ DefaultFetch::doSquash(const Addr &new_PC, delete retryPkt; } retryPkt = NULL; - retryTid = -1; + retryTid = InvalidThreadID; } fetchStatus[tid] = Squashing; @@ -702,7 +704,7 @@ template void DefaultFetch::squashFromDecode(const Addr &new_PC, const Addr &new_NPC, const Addr &new_MicroPC, - const InstSeqNum &seq_num, unsigned tid) + const InstSeqNum &seq_num, ThreadID tid) { DPRINTF(Fetch, "[tid:%i]: Squashing from decode.\n",tid); @@ -715,7 +717,7 @@ DefaultFetch::squashFromDecode(const Addr &new_PC, const Addr &new_NPC, template bool -DefaultFetch::checkStall(unsigned tid) const +DefaultFetch::checkStall(ThreadID tid) const { bool ret_val = false; @@ -744,11 +746,11 @@ typename DefaultFetch::FetchStatus DefaultFetch::updateFetchStatus() { //Check Running - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (fetchStatus[tid] == Running || fetchStatus[tid] == Squashing || @@ -783,7 +785,7 @@ template void DefaultFetch::squash(const Addr &new_PC, const Addr &new_NPC, const Addr &new_MicroPC, - const InstSeqNum &seq_num, unsigned tid) + const InstSeqNum &seq_num, ThreadID tid) { DPRINTF(Fetch, "[tid:%u]: Squash from commit.\n",tid); @@ -797,14 +799,14 @@ template void DefaultFetch::tick() { - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); bool status_change = false; wroteToTimeBuffer = false; while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; // Check the signals for each thread to determine the proper status // for each thread. @@ -851,7 +853,7 @@ DefaultFetch::tick() template bool -DefaultFetch::checkSignalsAndUpdate(unsigned tid) +DefaultFetch::checkSignalsAndUpdate(ThreadID tid) { // Update the per thread stall statuses. if (fromDecode->decodeBlock[tid]) { @@ -1000,9 +1002,9 @@ DefaultFetch::fetch(bool &status_change) ////////////////////////////////////////// // Start actual fetch ////////////////////////////////////////// - int tid = getFetchingThread(fetchPolicy); + ThreadID tid = getFetchingThread(fetchPolicy); - if (tid == -1 || drainPending) { + if (tid == InvalidThreadID || drainPending) { DPRINTF(Fetch,"There are no more threads available to fetch from.\n"); // Breaks looping condition in tick() @@ -1277,17 +1279,17 @@ DefaultFetch::recvRetry() { if (retryPkt != NULL) { assert(cacheBlocked); - assert(retryTid != -1); + assert(retryTid != InvalidThreadID); assert(fetchStatus[retryTid] == IcacheWaitRetry); if (icachePort->sendTiming(retryPkt)) { fetchStatus[retryTid] = IcacheWaitResponse; retryPkt = NULL; - retryTid = -1; + retryTid = InvalidThreadID; cacheBlocked = false; } } else { - assert(retryTid == -1); + assert(retryTid == InvalidThreadID); // Access has been squashed since it was sent out. Just clear // the cache being blocked. cacheBlocked = false; @@ -1300,7 +1302,7 @@ DefaultFetch::recvRetry() // // /////////////////////////////////////// template -int +ThreadID DefaultFetch::getFetchingThread(FetchPriority &fetch_priority) { if (numThreads > 1) { @@ -1322,36 +1324,35 @@ DefaultFetch::getFetchingThread(FetchPriority &fetch_priority) return branchCount(); default: - return -1; + return InvalidThreadID; } } else { - std::list::iterator thread = activeThreads->begin(); + list::iterator thread = activeThreads->begin(); if (thread == activeThreads->end()) { - return -1; + return InvalidThreadID; } - int tid = *thread; + ThreadID tid = *thread; if (fetchStatus[tid] == Running || fetchStatus[tid] == IcacheAccessComplete || fetchStatus[tid] == Idle) { return tid; } else { - return -1; + return InvalidThreadID; } } - } template -int +ThreadID DefaultFetch::roundRobin() { - std::list::iterator pri_iter = priorityList.begin(); - std::list::iterator end = priorityList.end(); + list::iterator pri_iter = priorityList.begin(); + list::iterator end = priorityList.end(); - int high_pri; + ThreadID high_pri; while (pri_iter != end) { high_pri = *pri_iter; @@ -1371,27 +1372,26 @@ DefaultFetch::roundRobin() pri_iter++; } - return -1; + return InvalidThreadID; } template -int +ThreadID DefaultFetch::iqCount() { - std::priority_queue PQ; + std::priority_queue PQ; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; PQ.push(fromIEW->iewInfo[tid].iqCount); } while (!PQ.empty()) { - - unsigned high_pri = PQ.top(); + ThreadID high_pri = PQ.top(); if (fetchStatus[high_pri] == Running || fetchStatus[high_pri] == IcacheAccessComplete || @@ -1402,27 +1402,26 @@ DefaultFetch::iqCount() } - return -1; + return InvalidThreadID; } template -int +ThreadID DefaultFetch::lsqCount() { - std::priority_queue PQ; + std::priority_queue PQ; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; PQ.push(fromIEW->iewInfo[tid].ldstqCount); } while (!PQ.empty()) { - - unsigned high_pri = PQ.top(); + ThreadID high_pri = PQ.top(); if (fetchStatus[high_pri] == Running || fetchStatus[high_pri] == IcacheAccessComplete || @@ -1430,20 +1429,21 @@ DefaultFetch::lsqCount() return high_pri; else PQ.pop(); - } - return -1; + return InvalidThreadID; } template -int +ThreadID DefaultFetch::branchCount() { - std::list::iterator thread = activeThreads->begin(); +#if 0 + list::iterator thread = activeThreads->begin(); assert(thread != activeThreads->end()); - unsigned tid = *thread; + ThreadID tid = *thread; +#endif panic("Branch Count Fetch policy unimplemented\n"); - return 0 * tid; + return InvalidThreadID; } diff --git a/src/cpu/o3/free_list.cc b/src/cpu/o3/free_list.cc index ae651398b..1144238f4 100644 --- a/src/cpu/o3/free_list.cc +++ b/src/cpu/o3/free_list.cc @@ -32,7 +32,7 @@ #include "cpu/o3/free_list.hh" -SimpleFreeList::SimpleFreeList(unsigned activeThreads, +SimpleFreeList::SimpleFreeList(ThreadID activeThreads, unsigned _numLogicalIntRegs, unsigned _numPhysicalIntRegs, unsigned _numLogicalFloatRegs, diff --git a/src/cpu/o3/free_list.hh b/src/cpu/o3/free_list.hh index 97b56909e..79e10524b 100644 --- a/src/cpu/o3/free_list.hh +++ b/src/cpu/o3/free_list.hh @@ -85,7 +85,7 @@ class SimpleFreeList * @param _numLogicalFloatRegs Number of logical fp registers. * @param _numPhysicalFloatRegs Number of physical fp registers. */ - SimpleFreeList(unsigned activeThreads, + SimpleFreeList(ThreadID activeThreads, unsigned _numLogicalIntRegs, unsigned _numPhysicalIntRegs, unsigned _numLogicalFloatRegs, diff --git a/src/cpu/o3/iew.hh b/src/cpu/o3/iew.hh index 3458f09d6..6797410d8 100644 --- a/src/cpu/o3/iew.hh +++ b/src/cpu/o3/iew.hh @@ -31,12 +31,11 @@ #ifndef __CPU_O3_IEW_HH__ #define __CPU_O3_IEW_HH__ -#include "config/full_system.hh" - #include #include "base/statistics.hh" #include "base/timebuf.hh" +#include "config/full_system.hh" #include "cpu/o3/comm.hh" #include "cpu/o3/scoreboard.hh" #include "cpu/o3/lsq.hh" @@ -139,7 +138,7 @@ class DefaultIEW void setIEWQueue(TimeBuffer *iq_ptr); /** Sets pointer to list of active threads. */ - void setActiveThreads(std::list *at_ptr); + void setActiveThreads(std::list *at_ptr); /** Sets pointer to the scoreboard. */ void setScoreboard(Scoreboard *sb_ptr); @@ -160,7 +159,7 @@ class DefaultIEW bool isSwitchedOut() { return switchedOut; } /** Squashes instructions in IEW for a specific thread. */ - void squash(unsigned tid); + void squash(ThreadID tid); /** Wakes all dependents of a completed instruction. */ void wakeDependents(DynInstPtr &inst); @@ -177,7 +176,7 @@ class DefaultIEW void instToCommit(DynInstPtr &inst); /** Inserts unused instructions of a thread into the skid buffer. */ - void skidInsert(unsigned tid); + void skidInsert(ThreadID tid); /** Returns the max of the number of entries in all of the skid buffers. */ int skidCount(); @@ -209,7 +208,7 @@ class DefaultIEW bool hasStoresToWB() { return ldstQueue.hasStoresToWB(); } /** Returns if the LSQ has any stores to writeback. */ - bool hasStoresToWB(unsigned tid) { return ldstQueue.hasStoresToWB(tid); } + bool hasStoresToWB(ThreadID tid) { return ldstQueue.hasStoresToWB(tid); } void incrWb(InstSeqNum &sn) { @@ -256,31 +255,31 @@ class DefaultIEW /** Sends commit proper information for a squash due to a branch * mispredict. */ - void squashDueToBranch(DynInstPtr &inst, unsigned thread_id); + void squashDueToBranch(DynInstPtr &inst, ThreadID tid); /** Sends commit proper information for a squash due to a memory order * violation. */ - void squashDueToMemOrder(DynInstPtr &inst, unsigned thread_id); + void squashDueToMemOrder(DynInstPtr &inst, ThreadID tid); /** Sends commit proper information for a squash due to memory becoming * blocked (younger issued instructions must be retried). */ - void squashDueToMemBlocked(DynInstPtr &inst, unsigned thread_id); + void squashDueToMemBlocked(DynInstPtr &inst, ThreadID tid); /** Sets Dispatch to blocked, and signals back to other stages to block. */ - void block(unsigned thread_id); + void block(ThreadID tid); /** Unblocks Dispatch if the skid buffer is empty, and signals back to * other stages to unblock. */ - void unblock(unsigned thread_id); + void unblock(ThreadID tid); /** Determines proper actions to take given Dispatch's status. */ - void dispatch(unsigned tid); + void dispatch(ThreadID tid); /** Dispatches instructions to IQ and LSQ. */ - void dispatchInsts(unsigned tid); + void dispatchInsts(ThreadID tid); /** Executes instructions. In the case of memory operations, it informs the * LSQ to execute the instructions. Also handles any redirects that occur @@ -301,16 +300,16 @@ class DefaultIEW unsigned validInstsFromRename(); /** Reads the stall signals. */ - void readStallSignals(unsigned tid); + void readStallSignals(ThreadID tid); /** Checks if any of the stall conditions are currently true. */ - bool checkStall(unsigned tid); + bool checkStall(ThreadID tid); /** Processes inputs and changes state accordingly. */ - void checkSignalsAndUpdate(unsigned tid); + void checkSignalsAndUpdate(ThreadID tid); /** Removes instructions from rename from a thread's instruction list. */ - void emptyRenameInsts(unsigned tid); + void emptyRenameInsts(ThreadID tid); /** Sorts instructions coming from rename into lists separated by thread. */ void sortInsts(); @@ -453,10 +452,10 @@ class DefaultIEW unsigned wbMax; /** Number of active threads. */ - unsigned numThreads; + ThreadID numThreads; /** Pointer to list of active threads. */ - std::list *activeThreads; + std::list *activeThreads; /** Maximum size of the skid buffer. */ unsigned skidBufferMax; diff --git a/src/cpu/o3/iew_impl.hh b/src/cpu/o3/iew_impl.hh index 778ff444f..ba29df196 100644 --- a/src/cpu/o3/iew_impl.hh +++ b/src/cpu/o3/iew_impl.hh @@ -39,6 +39,8 @@ #include "cpu/o3/iew.hh" #include "params/DerivO3CPU.hh" +using namespace std; + template DefaultIEW::DefaultIEW(O3CPU *_cpu, DerivO3CPUParams *params) : issueToExecQueue(params->backComSize, params->forwardComSize), @@ -66,10 +68,10 @@ DefaultIEW::DefaultIEW(O3CPU *_cpu, DerivO3CPUParams *params) // Instruction queue needs the queue between issue and execute. instQueue.setIssueToExecuteQueue(&issueToExecQueue); - for (int i=0; i < numThreads; i++) { - dispatchStatus[i] = Running; - stalls[i].commit = false; - fetchRedirect[i] = false; + for (ThreadID tid = 0; tid < numThreads; tid++) { + dispatchStatus[tid] = Running; + stalls[tid].commit = false; + fetchRedirect[tid] = false; } wbMax = wbWidth * params->wbDepth; @@ -164,7 +166,7 @@ DefaultIEW::regStats() .desc("Number of executed instructions"); iewExecLoadInsts - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".iewExecLoadInsts") .desc("Number of load instructions executed") .flags(total); @@ -174,25 +176,25 @@ DefaultIEW::regStats() .desc("Number of squashed instructions skipped in execute"); iewExecutedSwp - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".EXEC:swp") .desc("number of swp insts executed") .flags(total); iewExecutedNop - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".EXEC:nop") .desc("number of nop insts executed") .flags(total); iewExecutedRefs - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".EXEC:refs") .desc("number of memory reference insts executed") .flags(total); iewExecutedBranches - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".EXEC:branches") .desc("Number of branches executed") .flags(total); @@ -211,31 +213,31 @@ DefaultIEW::regStats() iewExecRate = iewExecutedInsts / cpu->numCycles; iewInstsToCommit - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:sent") .desc("cumulative count of insts sent to commit") .flags(total); writebackCount - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:count") .desc("cumulative count of insts written-back") .flags(total); producerInst - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:producers") .desc("num instructions producing a value") .flags(total); consumerInst - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:consumers") .desc("num instructions consuming a value") .flags(total); wbPenalized - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:penalized") .desc("number of instrctions required to write to 'other' IQ") .flags(total); @@ -265,7 +267,7 @@ template void DefaultIEW::initStage() { - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { toRename->iewInfo[tid].usedIQ = true; toRename->iewInfo[tid].freeIQEntries = instQueue.numFreeEntries(tid); @@ -318,7 +320,7 @@ DefaultIEW::setIEWQueue(TimeBuffer *iq_ptr) template void -DefaultIEW::setActiveThreads(std::list *at_ptr) +DefaultIEW::setActiveThreads(list *at_ptr) { activeThreads = at_ptr; @@ -361,11 +363,11 @@ DefaultIEW::switchOut() ldstQueue.switchOut(); fuPool->switchOut(); - for (int i = 0; i < numThreads; i++) { - while (!insts[i].empty()) - insts[i].pop(); - while (!skidBuffer[i].empty()) - skidBuffer[i].pop(); + for (ThreadID tid = 0; tid < numThreads; tid++) { + while (!insts[tid].empty()) + insts[tid].pop(); + while (!skidBuffer[tid].empty()) + skidBuffer[tid].pop(); } } @@ -386,10 +388,10 @@ DefaultIEW::takeOverFrom() initStage(); cpu->activityThisCycle(); - for (int i=0; i < numThreads; i++) { - dispatchStatus[i] = Running; - stalls[i].commit = false; - fetchRedirect[i] = false; + for (ThreadID tid = 0; tid < numThreads; tid++) { + dispatchStatus[tid] = Running; + stalls[tid].commit = false; + fetchRedirect[tid] = false; } updateLSQNextCycle = false; @@ -401,10 +403,9 @@ DefaultIEW::takeOverFrom() template void -DefaultIEW::squash(unsigned tid) +DefaultIEW::squash(ThreadID tid) { - DPRINTF(IEW, "[tid:%i]: Squashing all instructions.\n", - tid); + DPRINTF(IEW, "[tid:%i]: Squashing all instructions.\n", tid); // Tell the IQ to start squashing. instQueue.squash(tid); @@ -433,7 +434,7 @@ DefaultIEW::squash(unsigned tid) template void -DefaultIEW::squashDueToBranch(DynInstPtr &inst, unsigned tid) +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); @@ -464,7 +465,7 @@ DefaultIEW::squashDueToBranch(DynInstPtr &inst, unsigned tid) template void -DefaultIEW::squashDueToMemOrder(DynInstPtr &inst, unsigned tid) +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); @@ -482,7 +483,7 @@ DefaultIEW::squashDueToMemOrder(DynInstPtr &inst, unsigned tid) template void -DefaultIEW::squashDueToMemBlocked(DynInstPtr &inst, unsigned tid) +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); @@ -503,7 +504,7 @@ DefaultIEW::squashDueToMemBlocked(DynInstPtr &inst, unsigned tid) template void -DefaultIEW::block(unsigned tid) +DefaultIEW::block(ThreadID tid) { DPRINTF(IEW, "[tid:%u]: Blocking.\n", tid); @@ -522,7 +523,7 @@ DefaultIEW::block(unsigned tid) template void -DefaultIEW::unblock(unsigned tid) +DefaultIEW::unblock(ThreadID tid) { DPRINTF(IEW, "[tid:%i]: Reading instructions out of the skid " "buffer %u.\n",tid, tid); @@ -605,7 +606,7 @@ DefaultIEW::validInstsFromRename() template void -DefaultIEW::skidInsert(unsigned tid) +DefaultIEW::skidInsert(ThreadID tid) { DynInstPtr inst = NULL; @@ -631,11 +632,11 @@ DefaultIEW::skidCount() { int max=0; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; unsigned thread_count = skidBuffer[tid].size(); if (max < thread_count) max = thread_count; @@ -648,11 +649,11 @@ template bool DefaultIEW::skidsEmpty() { - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!skidBuffer[tid].empty()) return false; @@ -667,11 +668,11 @@ DefaultIEW::updateStatus() { bool any_unblocking = false; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (dispatchStatus[tid] == Unblocking) { any_unblocking = true; @@ -711,7 +712,7 @@ DefaultIEW::resetEntries() template void -DefaultIEW::readStallSignals(unsigned tid) +DefaultIEW::readStallSignals(ThreadID tid) { if (fromCommit->commitBlock[tid]) { stalls[tid].commit = true; @@ -725,7 +726,7 @@ DefaultIEW::readStallSignals(unsigned tid) template bool -DefaultIEW::checkStall(unsigned tid) +DefaultIEW::checkStall(ThreadID tid) { bool ret_val(false); @@ -761,7 +762,7 @@ DefaultIEW::checkStall(unsigned tid) template void -DefaultIEW::checkSignalsAndUpdate(unsigned tid) +DefaultIEW::checkSignalsAndUpdate(ThreadID tid) { // Check if there's a squash signal, squash if there is // Check stall signals, block if there is. @@ -834,8 +835,8 @@ DefaultIEW::sortInsts() { int insts_from_rename = fromRename->size; #ifdef DEBUG - for (int i = 0; i < numThreads; i++) - assert(insts[i].empty()); + for (ThreadID tid = 0; tid < numThreads; tid++) + assert(insts[tid].empty()); #endif for (int i = 0; i < insts_from_rename; ++i) { insts[fromRename->insts[i]->threadNumber].push(fromRename->insts[i]); @@ -844,7 +845,7 @@ DefaultIEW::sortInsts() template void -DefaultIEW::emptyRenameInsts(unsigned tid) +DefaultIEW::emptyRenameInsts(ThreadID tid) { DPRINTF(IEW, "[tid:%i]: Removing incoming rename instructions\n", tid); @@ -894,7 +895,7 @@ DefaultIEW::deactivateStage() template void -DefaultIEW::dispatch(unsigned tid) +DefaultIEW::dispatch(ThreadID tid) { // If status is Running or idle, // call dispatchInsts() @@ -942,7 +943,7 @@ DefaultIEW::dispatch(unsigned tid) template void -DefaultIEW::dispatchInsts(unsigned tid) +DefaultIEW::dispatchInsts(ThreadID tid) { // Obtain instructions from skid buffer if unblocking, or queue from rename // otherwise. @@ -1169,11 +1170,11 @@ DefaultIEW::executeInsts() wbNumInst = 0; wbCycle = 0; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; fetchRedirect[tid] = false; } @@ -1273,7 +1274,7 @@ DefaultIEW::executeInsts() // This probably needs to prioritize the redirects if a different // scheduler is used. Currently the scheduler schedules the oldest // instruction first, so the branch resolution order will be correct. - unsigned tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; if (!fetchRedirect[tid] || toCommit->squashedSeqNum[tid] > inst->seqNum) { @@ -1391,7 +1392,7 @@ DefaultIEW::writebackInsts() for (int inst_num = 0; inst_num < wbWidth && toCommit->insts[inst_num]; inst_num++) { DynInstPtr inst = toCommit->insts[inst_num]; - int tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; DPRINTF(IEW, "Sending instructions to commit, [sn:%lli] PC %#x.\n", inst->seqNum, inst->readPC()); @@ -1439,12 +1440,12 @@ DefaultIEW::tick() // Free function units marked as being freed this cycle. fuPool->processFreeUnits(); - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); // Check stall and squash signals, dispatch any instructions. while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; DPRINTF(IEW,"Issue: Processing [tid:%i]\n",tid); @@ -1486,7 +1487,7 @@ DefaultIEW::tick() threads = activeThreads->begin(); while (threads != end) { - unsigned tid = (*threads++); + ThreadID tid = (*threads++); DPRINTF(IEW,"Processing [tid:%i]\n",tid); @@ -1552,14 +1553,14 @@ template void DefaultIEW::updateExeInstStats(DynInstPtr &inst) { - int thread_number = inst->threadNumber; + ThreadID tid = inst->threadNumber; // // Pick off the software prefetches // #ifdef TARGET_ALPHA if (inst->isDataPrefetch()) - iewExecutedSwp[thread_number]++; + iewExecutedSwp[tid]++; else iewIewExecutedcutedInsts++; #else @@ -1570,16 +1571,16 @@ DefaultIEW::updateExeInstStats(DynInstPtr &inst) // Control operations // if (inst->isControl()) - iewExecutedBranches[thread_number]++; + iewExecutedBranches[tid]++; // // Memory operations // if (inst->isMemRef()) { - iewExecutedRefs[thread_number]++; + iewExecutedRefs[tid]++; if (inst->isLoad()) { - iewExecLoadInsts[thread_number]++; + iewExecLoadInsts[tid]++; } } } diff --git a/src/cpu/o3/inst_queue.hh b/src/cpu/o3/inst_queue.hh index 0b814ccb4..bccb1aa0e 100644 --- a/src/cpu/o3/inst_queue.hh +++ b/src/cpu/o3/inst_queue.hh @@ -126,7 +126,7 @@ class InstructionQueue void resetState(); /** Sets active threads list. */ - void setActiveThreads(std::list *at_ptr); + void setActiveThreads(std::list *at_ptr); /** Sets the timer buffer between issue and execute. */ void setIssueToExecuteQueue(TimeBuffer *i2eQueue); @@ -144,7 +144,7 @@ class InstructionQueue bool isSwitchedOut() { return switchedOut; } /** Number of entries needed for given amount of threads. */ - int entryAmount(int num_threads); + int entryAmount(ThreadID num_threads); /** Resets max entries for all threads. */ void resetEntries(); @@ -153,13 +153,13 @@ class InstructionQueue unsigned numFreeEntries(); /** Returns number of free entries for a thread. */ - unsigned numFreeEntries(unsigned tid); + unsigned numFreeEntries(ThreadID tid); /** Returns whether or not the IQ is full. */ bool isFull(); /** Returns whether or not the IQ is full for a specific thread. */ - bool isFull(unsigned tid); + bool isFull(ThreadID tid); /** Returns if there are any ready instructions in the IQ. */ bool hasReadyInsts(); @@ -203,7 +203,7 @@ class InstructionQueue * Commits all instructions up to and including the given sequence number, * for a specific thread. */ - void commit(const InstSeqNum &inst, unsigned tid = 0); + void commit(const InstSeqNum &inst, ThreadID tid = 0); /** Wakes all dependents of a completed instruction. */ int wakeDependents(DynInstPtr &completed_inst); @@ -230,17 +230,17 @@ class InstructionQueue * Squashes instructions for a thread. Squashing information is obtained * from the time buffer. */ - void squash(unsigned tid); + void squash(ThreadID tid); /** Returns the number of used entries for a thread. */ - unsigned getCount(unsigned tid) { return count[tid]; }; + unsigned getCount(ThreadID tid) { return count[tid]; }; /** Debug function to print all instructions. */ void printInsts(); private: /** Does the actual squashing. */ - void doSquash(unsigned tid); + void doSquash(ThreadID tid); ///////////////////////// // Various pointers @@ -368,10 +368,10 @@ class InstructionQueue IQPolicy iqPolicy; /** Number of Total Threads*/ - unsigned numThreads; + ThreadID numThreads; /** Pointer to list of active threads. */ - std::list *activeThreads; + std::list *activeThreads; /** Per Thread IQ count */ unsigned count[Impl::MaxThreads]; diff --git a/src/cpu/o3/inst_queue_impl.hh b/src/cpu/o3/inst_queue_impl.hh index 1d0f4b9f6..b6d1ec8b0 100644 --- a/src/cpu/o3/inst_queue_impl.hh +++ b/src/cpu/o3/inst_queue_impl.hh @@ -35,9 +35,10 @@ #include "cpu/o3/fu_pool.hh" #include "cpu/o3/inst_queue.hh" #include "enums/OpClass.hh" +#include "params/DerivO3CPU.hh" #include "sim/core.hh" -#include "params/DerivO3CPU.hh" +using namespace std; template InstructionQueue::FUCompletion::FUCompletion(DynInstPtr &_inst, @@ -93,9 +94,9 @@ InstructionQueue::InstructionQueue(O3CPU *cpu_ptr, IEW *iew_ptr, regScoreboard.resize(numPhysRegs); //Initialize Mem Dependence Units - for (int i = 0; i < numThreads; i++) { - memDepUnit[i].init(params,i); - memDepUnit[i].setIQ(this); + for (ThreadID tid = 0; tid < numThreads; tid++) { + memDepUnit[tid].init(params, tid); + memDepUnit[tid].setIQ(this); } resetState(); @@ -111,8 +112,8 @@ InstructionQueue::InstructionQueue(O3CPU *cpu_ptr, IEW *iew_ptr, iqPolicy = Dynamic; //Set Max Entries to Total ROB Capacity - for (int i = 0; i < numThreads; i++) { - maxEntries[i] = numEntries; + for (ThreadID tid = 0; tid < numThreads; tid++) { + maxEntries[tid] = numEntries; } } else if (policy == "partitioned") { @@ -122,8 +123,8 @@ InstructionQueue::InstructionQueue(O3CPU *cpu_ptr, IEW *iew_ptr, int part_amt = numEntries / numThreads; //Divide ROB up evenly - for (int i = 0; i < numThreads; i++) { - maxEntries[i] = part_amt; + for (ThreadID tid = 0; tid < numThreads; tid++) { + maxEntries[tid] = part_amt; } DPRINTF(IQ, "IQ sharing policy set to Partitioned:" @@ -136,8 +137,8 @@ InstructionQueue::InstructionQueue(O3CPU *cpu_ptr, IEW *iew_ptr, int thresholdIQ = (int)((double)threshold * numEntries); //Divide up by threshold amount - for (int i = 0; i < numThreads; i++) { - maxEntries[i] = thresholdIQ; + for (ThreadID tid = 0; tid < numThreads; tid++) { + maxEntries[tid] = thresholdIQ; } DPRINTF(IQ, "IQ sharing policy set to Threshold:" @@ -315,9 +316,9 @@ InstructionQueue::regStats() ; fuBusyRate = fuBusy / iqInstsIssued; - for ( int i=0; i < numThreads; i++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { // Tell mem dependence unit to reg stats as well. - memDepUnit[i].regStats(); + memDepUnit[tid].regStats(); } } @@ -326,9 +327,9 @@ void InstructionQueue::resetState() { //Initialize thread IQ counts - for (int i = 0; i ::resetState() regScoreboard[i] = false; } - for (int i = 0; i < numThreads; ++i) { - squashedSeqNum[i] = 0; + for (ThreadID tid = 0; tid < numThreads; ++tid) { + squashedSeqNum[tid] = 0; } for (int i = 0; i < Num_OpClasses; ++i) { @@ -359,7 +360,7 @@ InstructionQueue::resetState() template void -InstructionQueue::setActiveThreads(std::list *at_ptr) +InstructionQueue::setActiveThreads(list *at_ptr) { activeThreads = at_ptr; } @@ -395,8 +396,8 @@ InstructionQueue::switchOut() dependGraph.reset(); instsToExecute.clear(); switchedOut = true; - for (int i = 0; i < numThreads; ++i) { - memDepUnit[i].switchOut(); + for (ThreadID tid = 0; tid < numThreads; ++tid) { + memDepUnit[tid].switchOut(); } } @@ -409,7 +410,7 @@ InstructionQueue::takeOverFrom() template int -InstructionQueue::entryAmount(int num_threads) +InstructionQueue::entryAmount(ThreadID num_threads) { if (iqPolicy == Partitioned) { return numEntries / num_threads; @@ -426,11 +427,11 @@ InstructionQueue::resetEntries() if (iqPolicy != Dynamic || numThreads > 1) { int active_threads = activeThreads->size(); - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (iqPolicy == Partitioned) { maxEntries[tid] = numEntries / active_threads; @@ -450,7 +451,7 @@ InstructionQueue::numFreeEntries() template unsigned -InstructionQueue::numFreeEntries(unsigned tid) +InstructionQueue::numFreeEntries(ThreadID tid) { return maxEntries[tid] - count[tid]; } @@ -470,7 +471,7 @@ InstructionQueue::isFull() template bool -InstructionQueue::isFull(unsigned tid) +InstructionQueue::isFull(ThreadID tid) { if (numFreeEntries(tid) == 0) { return(true); @@ -726,7 +727,7 @@ InstructionQueue::scheduleReadyInsts() int idx = -2; int op_latency = 1; - int tid = issuing_inst->threadNumber; + ThreadID tid = issuing_inst->threadNumber; if (op_class != No_OpClass) { idx = fuPool->getUnit(op_class); @@ -825,7 +826,7 @@ InstructionQueue::scheduleNonSpec(const InstSeqNum &inst) assert(inst_it != nonSpecInsts.end()); - unsigned tid = (*inst_it).second->threadNumber; + ThreadID tid = (*inst_it).second->threadNumber; (*inst_it).second->setAtCommit(); @@ -844,7 +845,7 @@ InstructionQueue::scheduleNonSpec(const InstSeqNum &inst) template void -InstructionQueue::commit(const InstSeqNum &inst, unsigned tid) +InstructionQueue::commit(const InstSeqNum &inst, ThreadID tid) { DPRINTF(IQ, "[tid:%i]: Committing instructions older than [sn:%i]\n", tid,inst); @@ -976,7 +977,7 @@ template void InstructionQueue::completeMemInst(DynInstPtr &completed_inst) { - int tid = completed_inst->threadNumber; + ThreadID tid = completed_inst->threadNumber; DPRINTF(IQ, "Completing mem instruction PC:%#x [sn:%lli]\n", completed_inst->readPC(), completed_inst->seqNum); @@ -999,7 +1000,7 @@ InstructionQueue::violation(DynInstPtr &store, template void -InstructionQueue::squash(unsigned tid) +InstructionQueue::squash(ThreadID tid) { DPRINTF(IQ, "[tid:%i]: Starting to squash instructions in " "the IQ.\n", tid); @@ -1019,7 +1020,7 @@ InstructionQueue::squash(unsigned tid) template void -InstructionQueue::doSquash(unsigned tid) +InstructionQueue::doSquash(ThreadID tid) { // Start at the tail. ListIt squash_it = instList[tid].end(); @@ -1253,10 +1254,10 @@ InstructionQueue::countInsts() // Change the #if if you want to use this method. int total_insts = 0; - for (int i = 0; i < numThreads; ++i) { - ListIt count_it = instList[i].begin(); + for (ThreadID tid = 0; tid < numThreads; ++tid) { + ListIt count_it = instList[tid].begin(); - while (count_it != instList[i].end()) { + while (count_it != instList[tid].end()) { if (!(*count_it)->isSquashed() && !(*count_it)->isSquashedInIQ()) { if (!(*count_it)->isIssued()) { ++total_insts; @@ -1325,15 +1326,13 @@ template void InstructionQueue::dumpInsts() { - for (int i = 0; i < numThreads; ++i) { + for (ThreadID tid = 0; tid < numThreads; ++tid) { int num = 0; int valid_num = 0; - ListIt inst_list_it = instList[i].begin(); + ListIt inst_list_it = instList[tid].begin(); - while (inst_list_it != instList[i].end()) - { - cprintf("Instruction:%i\n", - num); + while (inst_list_it != instList[tid].end()) { + cprintf("Instruction:%i\n", num); if (!(*inst_list_it)->isSquashed()) { if (!(*inst_list_it)->isIssued()) { ++valid_num; diff --git a/src/cpu/o3/lsq.hh b/src/cpu/o3/lsq.hh index cf27552d4..a0bae058c 100644 --- a/src/cpu/o3/lsq.hh +++ b/src/cpu/o3/lsq.hh @@ -74,24 +74,24 @@ class LSQ { Port *getDcachePort() { return &dcachePort; } /** Sets the pointer to the list of active threads. */ - void setActiveThreads(std::list *at_ptr); + void setActiveThreads(std::list *at_ptr); /** Switches out the LSQ. */ void switchOut(); /** Takes over execution from another CPU's thread. */ void takeOverFrom(); /** Number of entries needed for the given amount of threads.*/ - int entryAmount(int num_threads); - void removeEntries(unsigned tid); + int entryAmount(ThreadID num_threads); + void removeEntries(ThreadID tid); /** Reset the max entries for each thread. */ void resetEntries(); /** Resize the max entries for a thread. */ - void resizeEntries(unsigned size, unsigned tid); + void resizeEntries(unsigned size, ThreadID tid); /** Ticks the LSQ. */ void tick(); /** Ticks a specific LSQ Unit. */ - void tick(unsigned tid) + void tick(ThreadID tid) { thread[tid].tick(); } /** Inserts a load into the LSQ. */ @@ -108,13 +108,13 @@ class LSQ { /** * Commits loads up until the given sequence number for a specific thread. */ - void commitLoads(InstSeqNum &youngest_inst, unsigned tid) + void commitLoads(InstSeqNum &youngest_inst, ThreadID tid) { thread[tid].commitLoads(youngest_inst); } /** * Commits stores up until the given sequence number for a specific thread. */ - void commitStores(InstSeqNum &youngest_inst, unsigned tid) + void commitStores(InstSeqNum &youngest_inst, ThreadID tid) { thread[tid].commitStores(youngest_inst); } /** @@ -123,12 +123,12 @@ class LSQ { */ void writebackStores(); /** Same as above, but only for one thread. */ - void writebackStores(unsigned tid); + void writebackStores(ThreadID tid); /** * Squash instructions from a thread until the specified sequence number. */ - void squash(const InstSeqNum &squashed_num, unsigned tid) + void squash(const InstSeqNum &squashed_num, ThreadID tid) { thread[tid].squash(squashed_num); } /** Returns whether or not there was a memory ordering violation. */ @@ -137,41 +137,41 @@ class LSQ { * Returns whether or not there was a memory ordering violation for a * specific thread. */ - bool violation(unsigned tid) + bool violation(ThreadID tid) { return thread[tid].violation(); } /** Returns if a load is blocked due to the memory system for a specific * thread. */ - bool loadBlocked(unsigned tid) + bool loadBlocked(ThreadID tid) { return thread[tid].loadBlocked(); } - bool isLoadBlockedHandled(unsigned tid) + bool isLoadBlockedHandled(ThreadID tid) { return thread[tid].isLoadBlockedHandled(); } - void setLoadBlockedHandled(unsigned tid) + void setLoadBlockedHandled(ThreadID tid) { thread[tid].setLoadBlockedHandled(); } /** Gets the instruction that caused the memory ordering violation. */ - DynInstPtr getMemDepViolator(unsigned tid) + DynInstPtr getMemDepViolator(ThreadID tid) { return thread[tid].getMemDepViolator(); } /** Returns the head index of the load queue for a specific thread. */ - int getLoadHead(unsigned tid) + int getLoadHead(ThreadID tid) { return thread[tid].getLoadHead(); } /** Returns the sequence number of the head of the load queue. */ - InstSeqNum getLoadHeadSeqNum(unsigned tid) + InstSeqNum getLoadHeadSeqNum(ThreadID tid) { return thread[tid].getLoadHeadSeqNum(); } /** Returns the head index of the store queue. */ - int getStoreHead(unsigned tid) + int getStoreHead(ThreadID tid) { return thread[tid].getStoreHead(); } /** Returns the sequence number of the head of the store queue. */ - InstSeqNum getStoreHeadSeqNum(unsigned tid) + InstSeqNum getStoreHeadSeqNum(ThreadID tid) { return thread[tid].getStoreHeadSeqNum(); } @@ -179,31 +179,31 @@ class LSQ { /** Returns the number of instructions in all of the queues. */ int getCount(); /** Returns the number of instructions in the queues of one thread. */ - int getCount(unsigned tid) + int getCount(ThreadID tid) { return thread[tid].getCount(); } /** Returns the total number of loads in the load queue. */ int numLoads(); /** Returns the total number of loads for a single thread. */ - int numLoads(unsigned tid) + int numLoads(ThreadID tid) { return thread[tid].numLoads(); } /** Returns the total number of stores in the store queue. */ int numStores(); /** Returns the total number of stores for a single thread. */ - int numStores(unsigned tid) + int numStores(ThreadID tid) { return thread[tid].numStores(); } /** Returns the total number of loads that are ready. */ int numLoadsReady(); /** Returns the number of loads that are ready for a single thread. */ - int numLoadsReady(unsigned tid) + int numLoadsReady(ThreadID tid) { return thread[tid].numLoadsReady(); } /** Returns the number of free entries. */ unsigned numFreeEntries(); /** Returns the number of free entries for a specific thread. */ - unsigned numFreeEntries(unsigned tid); + unsigned numFreeEntries(ThreadID tid); /** Returns if the LSQ is full (either LQ or SQ is full). */ bool isFull(); @@ -211,17 +211,17 @@ class LSQ { * Returns if the LSQ is full for a specific thread (either LQ or SQ is * full). */ - bool isFull(unsigned tid); + bool isFull(ThreadID tid); /** Returns if any of the LQs are full. */ bool lqFull(); /** Returns if the LQ of a given thread is full. */ - bool lqFull(unsigned tid); + bool lqFull(ThreadID tid); /** Returns if any of the SQs are full. */ bool sqFull(); /** Returns if the SQ of a given thread is full. */ - bool sqFull(unsigned tid); + bool sqFull(ThreadID tid); /** * Returns if the LSQ is stalled due to a memory operation that must be @@ -232,7 +232,7 @@ class LSQ { * Returns if the LSQ of a specific thread is stalled due to a memory * operation that must be replayed. */ - bool isStalled(unsigned tid); + bool isStalled(ThreadID tid); /** Returns whether or not there are any stores to write back to memory. */ bool hasStoresToWB(); @@ -240,11 +240,11 @@ class LSQ { /** Returns whether or not a specific thread has any stores to write back * to memory. */ - bool hasStoresToWB(unsigned tid) + bool hasStoresToWB(ThreadID tid) { return thread[tid].hasStoresToWB(); } /** Returns the number of stores a specific thread has to write back. */ - int numStoresToWB(unsigned tid) + int numStoresToWB(ThreadID tid) { return thread[tid].numStoresToWB(); } /** Returns if the LSQ will write back to memory this cycle. */ @@ -252,22 +252,22 @@ class LSQ { /** Returns if the LSQ of a specific thread will write back to memory this * cycle. */ - bool willWB(unsigned tid) + bool willWB(ThreadID tid) { return thread[tid].willWB(); } /** Returns if the cache is currently blocked. */ bool cacheBlocked() - { return retryTid != -1; } + { return retryTid != InvalidThreadID; } /** Sets the retry thread id, indicating that one of the LSQUnits * tried to access the cache but the cache was blocked. */ - void setRetryTid(int tid) + void setRetryTid(ThreadID tid) { retryTid = tid; } /** Debugging function to print out all instructions. */ void dumpInsts(); /** Debugging function to print out instructions from a specific thread. */ - void dumpInsts(unsigned tid) + void dumpInsts(ThreadID tid) { thread[tid].dumpInsts(); } /** Executes a read operation, using the load specified at the load index. */ @@ -345,7 +345,7 @@ class LSQ { LSQUnit thread[Impl::MaxThreads]; /** List of Active Threads in System. */ - std::list *activeThreads; + std::list *activeThreads; /** Total Size of LQ Entries. */ unsigned LQEntries; @@ -359,11 +359,11 @@ class LSQ { unsigned maxSQEntries; /** Number of Threads. */ - unsigned numThreads; + ThreadID numThreads; /** The thread id of the LSQ Unit that is currently waiting for a * retry. */ - int retryTid; + ThreadID retryTid; }; template @@ -371,7 +371,7 @@ template Fault LSQ::read(RequestPtr req, T &data, int load_idx) { - unsigned tid = req->threadId(); + ThreadID tid = req->threadId(); return thread[tid].read(req, data, load_idx); } @@ -381,7 +381,7 @@ template Fault LSQ::write(RequestPtr req, T &data, int store_idx) { - unsigned tid = req->threadId(); + ThreadID tid = req->threadId(); return thread[tid].write(req, data, store_idx); } diff --git a/src/cpu/o3/lsq_impl.hh b/src/cpu/o3/lsq_impl.hh index 8f9f63081..e780b14e4 100644 --- a/src/cpu/o3/lsq_impl.hh +++ b/src/cpu/o3/lsq_impl.hh @@ -33,9 +33,10 @@ #include #include "cpu/o3/lsq.hh" - #include "params/DerivO3CPU.hh" +using namespace std; + template void LSQ::DcachePort::setPeer(Port *port) @@ -170,7 +171,7 @@ LSQ::LSQ(O3CPU *cpu_ptr, IEW *iew_ptr, DerivO3CPUParams *params) } //Initialize LSQs - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { thread[tid].init(cpu, iew_ptr, params, this, maxLQEntries, maxSQEntries, tid); thread[tid].setDcachePort(&dcachePort); @@ -190,14 +191,14 @@ void LSQ::regStats() { //Initialize LSQs - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { thread[tid].regStats(); } } template void -LSQ::setActiveThreads(std::list *at_ptr) +LSQ::setActiveThreads(list *at_ptr) { activeThreads = at_ptr; assert(activeThreads != 0); @@ -207,7 +208,7 @@ template void LSQ::switchOut() { - for (int tid = 0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { thread[tid].switchOut(); } } @@ -216,14 +217,14 @@ template void LSQ::takeOverFrom() { - for (int tid = 0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { thread[tid].takeOverFrom(); } } template int -LSQ::entryAmount(int num_threads) +LSQ::entryAmount(ThreadID num_threads) { if (lsqPolicy == Partitioned) { return LQEntries / num_threads; @@ -249,11 +250,11 @@ LSQ::resetEntries() maxEntries = LQEntries; } - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; resizeEntries(maxEntries, tid); } @@ -262,7 +263,7 @@ LSQ::resetEntries() template void -LSQ::removeEntries(unsigned tid) +LSQ::removeEntries(ThreadID tid) { thread[tid].clearLQ(); thread[tid].clearSQ(); @@ -270,7 +271,7 @@ LSQ::removeEntries(unsigned tid) template void -LSQ::resizeEntries(unsigned size,unsigned tid) +LSQ::resizeEntries(unsigned size, ThreadID tid) { thread[tid].resizeLQ(size); thread[tid].resizeSQ(size); @@ -280,11 +281,11 @@ template void LSQ::tick() { - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; thread[tid].tick(); } @@ -294,7 +295,7 @@ template void LSQ::insertLoad(DynInstPtr &load_inst) { - unsigned tid = load_inst->threadNumber; + ThreadID tid = load_inst->threadNumber; thread[tid].insertLoad(load_inst); } @@ -303,7 +304,7 @@ template void LSQ::insertStore(DynInstPtr &store_inst) { - unsigned tid = store_inst->threadNumber; + ThreadID tid = store_inst->threadNumber; thread[tid].insertStore(store_inst); } @@ -312,7 +313,7 @@ template Fault LSQ::executeLoad(DynInstPtr &inst) { - unsigned tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; return thread[tid].executeLoad(inst); } @@ -321,7 +322,7 @@ template Fault LSQ::executeStore(DynInstPtr &inst) { - unsigned tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; return thread[tid].executeStore(inst); } @@ -330,11 +331,11 @@ template void LSQ::writebackStores() { - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (numStoresToWB(tid) > 0) { DPRINTF(Writeback,"[tid:%i] Writing back stores. %i stores " @@ -350,11 +351,11 @@ bool LSQ::violation() { /* Answers: Does Anybody Have a Violation?*/ - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (thread[tid].violation()) return true; @@ -369,11 +370,11 @@ LSQ::getCount() { unsigned total = 0; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; total += getCount(tid); } @@ -387,11 +388,11 @@ LSQ::numLoads() { unsigned total = 0; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; total += numLoads(tid); } @@ -405,11 +406,11 @@ LSQ::numStores() { unsigned total = 0; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; total += thread[tid].numStores(); } @@ -423,11 +424,11 @@ LSQ::numLoadsReady() { unsigned total = 0; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; total += thread[tid].numLoadsReady(); } @@ -441,11 +442,11 @@ LSQ::numFreeEntries() { unsigned total = 0; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; total += thread[tid].numFreeEntries(); } @@ -455,7 +456,7 @@ LSQ::numFreeEntries() template unsigned -LSQ::numFreeEntries(unsigned tid) +LSQ::numFreeEntries(ThreadID tid) { //if (lsqPolicy == Dynamic) //return numFreeEntries(); @@ -467,11 +468,11 @@ template bool LSQ::isFull() { - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!(thread[tid].lqFull() || thread[tid].sqFull())) return false; @@ -482,7 +483,7 @@ LSQ::isFull() template bool -LSQ::isFull(unsigned tid) +LSQ::isFull(ThreadID tid) { //@todo: Change to Calculate All Entries for //Dynamic Policy @@ -496,11 +497,11 @@ template bool LSQ::lqFull() { - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!thread[tid].lqFull()) return false; @@ -511,7 +512,7 @@ LSQ::lqFull() template bool -LSQ::lqFull(unsigned tid) +LSQ::lqFull(ThreadID tid) { //@todo: Change to Calculate All Entries for //Dynamic Policy @@ -525,11 +526,11 @@ template bool LSQ::sqFull() { - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!sqFull(tid)) return false; @@ -540,7 +541,7 @@ LSQ::sqFull() template bool -LSQ::sqFull(unsigned tid) +LSQ::sqFull(ThreadID tid) { //@todo: Change to Calculate All Entries for //Dynamic Policy @@ -554,11 +555,11 @@ template bool LSQ::isStalled() { - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!thread[tid].isStalled()) return false; @@ -569,7 +570,7 @@ LSQ::isStalled() template bool -LSQ::isStalled(unsigned tid) +LSQ::isStalled(ThreadID tid) { if (lsqPolicy == Dynamic) return isStalled(); @@ -581,11 +582,11 @@ template bool LSQ::hasStoresToWB() { - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (hasStoresToWB(tid)) return true; @@ -598,11 +599,11 @@ template bool LSQ::willWB() { - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (willWB(tid)) return true; @@ -615,11 +616,11 @@ template void LSQ::dumpInsts() { - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; thread[tid].dumpInsts(); } diff --git a/src/cpu/o3/lsq_unit.hh b/src/cpu/o3/lsq_unit.hh index 28e6f4506..a917caef3 100644 --- a/src/cpu/o3/lsq_unit.hh +++ b/src/cpu/o3/lsq_unit.hh @@ -325,7 +325,7 @@ class LSQUnit { private: /** The LSQUnit thread id. */ - unsigned lsqID; + ThreadID lsqID; /** The store queue. */ std::vector storeQueue; diff --git a/src/cpu/o3/lsq_unit_impl.hh b/src/cpu/o3/lsq_unit_impl.hh index afc9faf9c..edc8c9b3f 100644 --- a/src/cpu/o3/lsq_unit_impl.hh +++ b/src/cpu/o3/lsq_unit_impl.hh @@ -938,7 +938,7 @@ LSQUnit::recvRetry() storePostSend(retryPkt); retryPkt = NULL; isStoreBlocked = false; - lsq->setRetryTid(-1); + lsq->setRetryTid(InvalidThreadID); } else { // Still blocked! ++lsqCacheBlocked; diff --git a/src/cpu/o3/mem_dep_unit.hh b/src/cpu/o3/mem_dep_unit.hh index 4f9e7c9f7..a9560f446 100644 --- a/src/cpu/o3/mem_dep_unit.hh +++ b/src/cpu/o3/mem_dep_unit.hh @@ -86,7 +86,7 @@ class MemDepUnit std::string name() const { return _name; } /** Initializes the unit with parameters and a thread id. */ - void init(DerivO3CPUParams *params, int tid); + void init(DerivO3CPUParams *params, ThreadID tid); /** Registers statistics. */ void regStats(); @@ -135,7 +135,7 @@ class MemDepUnit /** Squashes all instructions up until a given sequence number for a * specific thread. */ - void squash(const InstSeqNum &squashed_num, unsigned tid); + void squash(const InstSeqNum &squashed_num, ThreadID tid); /** Indicates an ordering violation between a store and a younger load. */ void violation(DynInstPtr &store_inst, DynInstPtr &violating_load); diff --git a/src/cpu/o3/mem_dep_unit_impl.hh b/src/cpu/o3/mem_dep_unit_impl.hh index 8754539f9..59344d9f7 100644 --- a/src/cpu/o3/mem_dep_unit_impl.hh +++ b/src/cpu/o3/mem_dep_unit_impl.hh @@ -54,7 +54,7 @@ MemDepUnit::MemDepUnit(DerivO3CPUParams *params) template MemDepUnit::~MemDepUnit() { - for (int tid=0; tid < Impl::MaxThreads; tid++) { + for (ThreadID tid = 0; tid < Impl::MaxThreads; tid++) { ListIt inst_list_it = instList[tid].begin(); @@ -78,7 +78,7 @@ MemDepUnit::~MemDepUnit() template void -MemDepUnit::init(DerivO3CPUParams *params, int tid) +MemDepUnit::init(DerivO3CPUParams *params, ThreadID tid) { DPRINTF(MemDepUnit, "Creating MemDepUnit %i object.\n",tid); @@ -145,7 +145,7 @@ template void MemDepUnit::insert(DynInstPtr &inst) { - unsigned tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; MemDepEntryPtr inst_entry = new MemDepEntry(inst); @@ -242,7 +242,7 @@ template void MemDepUnit::insertNonSpec(DynInstPtr &inst) { - unsigned tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; MemDepEntryPtr inst_entry = new MemDepEntry(inst); @@ -292,7 +292,7 @@ MemDepUnit::insertBarrier(DynInstPtr &barr_inst) DPRINTF(MemDepUnit, "Inserted a write barrier\n"); } - unsigned tid = barr_inst->threadNumber; + ThreadID tid = barr_inst->threadNumber; MemDepEntryPtr inst_entry = new MemDepEntry(barr_inst); @@ -382,7 +382,7 @@ MemDepUnit::completed(DynInstPtr &inst) "[sn:%lli].\n", inst->readPC(), inst->seqNum); - unsigned tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; // Remove the instruction from the hash and the list. MemDepHashIt hash_it = memDepHash.find(inst->seqNum); @@ -457,7 +457,7 @@ MemDepUnit::wakeDependents(DynInstPtr &inst) template void MemDepUnit::squash(const InstSeqNum &squashed_num, - unsigned tid) + ThreadID tid) { if (!instsToReplay.empty()) { ListIt replay_it = instsToReplay.begin(); @@ -552,7 +552,7 @@ template void MemDepUnit::dumpLists() { - for (unsigned tid=0; tid < Impl::MaxThreads; tid++) { + for (ThreadID tid = 0; tid < Impl::MaxThreads; tid++) { cprintf("Instruction list %i size: %i\n", tid, instList[tid].size()); diff --git a/src/cpu/o3/regfile.hh b/src/cpu/o3/regfile.hh index 53ac2d683..07f8d487b 100644 --- a/src/cpu/o3/regfile.hh +++ b/src/cpu/o3/regfile.hh @@ -230,27 +230,28 @@ class PhysRegFile floatRegFile[reg_idx].q = val; } - MiscReg readMiscRegNoEffect(int misc_reg, unsigned thread_id) + MiscReg + readMiscRegNoEffect(int misc_reg, ThreadID tid) { - return miscRegs[thread_id].readRegNoEffect(misc_reg); + return miscRegs[tid].readRegNoEffect(misc_reg); } - MiscReg readMiscReg(int misc_reg, unsigned thread_id) + MiscReg + readMiscReg(int misc_reg, ThreadID tid) { - return miscRegs[thread_id].readReg(misc_reg, cpu->tcBase(thread_id)); + return miscRegs[tid].readReg(misc_reg, cpu->tcBase(tid)); } - void setMiscRegNoEffect(int misc_reg, - const MiscReg &val, unsigned thread_id) + void + setMiscRegNoEffect(int misc_reg, const MiscReg &val, ThreadID tid) { - miscRegs[thread_id].setRegNoEffect(misc_reg, val); + miscRegs[tid].setRegNoEffect(misc_reg, val); } - void setMiscReg(int misc_reg, const MiscReg &val, - unsigned thread_id) + void + setMiscReg(int misc_reg, const MiscReg &val, ThreadID tid) { - miscRegs[thread_id].setReg(misc_reg, val, - cpu->tcBase(thread_id)); + miscRegs[tid].setReg(misc_reg, val, cpu->tcBase(tid)); } public: diff --git a/src/cpu/o3/rename.hh b/src/cpu/o3/rename.hh index 0fdf28b19..734b63105 100644 --- a/src/cpu/o3/rename.hh +++ b/src/cpu/o3/rename.hh @@ -145,7 +145,7 @@ class DefaultRename void initStage(); /** Sets pointer to list of active threads. */ - void setActiveThreads(std::list *at_ptr); + void setActiveThreads(std::list *at_ptr); /** Sets pointer to rename maps (per-thread structures). */ void setRenameMap(RenameMap rm_ptr[Impl::MaxThreads]); @@ -169,7 +169,7 @@ class DefaultRename void takeOverFrom(); /** Squashes all instructions in a thread. */ - void squash(const InstSeqNum &squash_seq_num, unsigned tid); + void squash(const InstSeqNum &squash_seq_num, ThreadID tid); /** Ticks rename, which processes all input signals and attempts to rename * as many instructions as possible. @@ -185,17 +185,17 @@ class DefaultRename * change (ie switching from blocking to unblocking). * @param tid Thread id to rename instructions from. */ - void rename(bool &status_change, unsigned tid); + void rename(bool &status_change, ThreadID tid); /** Renames instructions for the given thread. Also handles serializing * instructions. */ - void renameInsts(unsigned tid); + void renameInsts(ThreadID tid); /** Inserts unused instructions from a given thread into the skid buffer, * to be renamed once rename unblocks. */ - void skidInsert(unsigned tid); + void skidInsert(ThreadID tid); /** Separates instructions from decode into individual lists of instructions * sorted by thread. @@ -212,49 +212,49 @@ class DefaultRename * blocked. * @return Returns true if there is a status change. */ - bool block(unsigned tid); + bool block(ThreadID tid); /** Switches rename to unblocking if the skid buffer is empty, and signals * back that rename has unblocked. * @return Returns true if there is a status change. */ - bool unblock(unsigned tid); + bool unblock(ThreadID tid); /** Executes actual squash, removing squashed instructions. */ - void doSquash(const InstSeqNum &squash_seq_num, unsigned tid); + void doSquash(const InstSeqNum &squash_seq_num, ThreadID tid); /** Removes a committed instruction's rename history. */ - void removeFromHistory(InstSeqNum inst_seq_num, unsigned tid); + void removeFromHistory(InstSeqNum inst_seq_num, ThreadID tid); /** Renames the source registers of an instruction. */ - inline void renameSrcRegs(DynInstPtr &inst, unsigned tid); + inline void renameSrcRegs(DynInstPtr &inst, ThreadID tid); /** Renames the destination registers of an instruction. */ - inline void renameDestRegs(DynInstPtr &inst, unsigned tid); + inline void renameDestRegs(DynInstPtr &inst, ThreadID tid); /** Calculates the number of free ROB entries for a specific thread. */ - inline int calcFreeROBEntries(unsigned tid); + inline int calcFreeROBEntries(ThreadID tid); /** Calculates the number of free IQ entries for a specific thread. */ - inline int calcFreeIQEntries(unsigned tid); + inline int calcFreeIQEntries(ThreadID tid); /** Calculates the number of free LSQ entries for a specific thread. */ - inline int calcFreeLSQEntries(unsigned tid); + inline int calcFreeLSQEntries(ThreadID tid); /** Returns the number of valid instructions coming from decode. */ unsigned validInsts(); /** Reads signals telling rename to block/unblock. */ - void readStallSignals(unsigned tid); + void readStallSignals(ThreadID tid); /** Checks if any stages are telling rename to block. */ - bool checkStall(unsigned tid); + bool checkStall(ThreadID tid); /** Gets the number of free entries for a specific thread. */ - void readFreeEntries(unsigned tid); + void readFreeEntries(ThreadID tid); /** Checks the signals and updates the status. */ - bool checkSignalsAndUpdate(unsigned tid); + bool checkSignalsAndUpdate(ThreadID tid); /** Either serializes on the next instruction available in the InstQueue, * or records that it must serialize on the next instruction to enter @@ -263,7 +263,7 @@ class DefaultRename * thread that has the serializeAfter instruction. * @param tid The thread id. */ - void serializeAfter(InstQueue &inst_list, unsigned tid); + void serializeAfter(InstQueue &inst_list, ThreadID tid); /** Holds the information for each destination register rename. It holds * the instruction's sequence number, the arch register, the old physical @@ -332,7 +332,7 @@ class DefaultRename FreeList *freeList; /** Pointer to the list of active threads. */ - std::list *activeThreads; + std::list *activeThreads; /** Pointer to the scoreboard. */ Scoreboard *scoreboard; @@ -418,7 +418,7 @@ class DefaultRename bool resumeUnblocking; /** The number of threads active in rename. */ - unsigned numThreads; + ThreadID numThreads; /** The maximum skid buffer size. */ unsigned skidBufferMax; diff --git a/src/cpu/o3/rename_impl.hh b/src/cpu/o3/rename_impl.hh index 81647b133..2bca6f81c 100644 --- a/src/cpu/o3/rename_impl.hh +++ b/src/cpu/o3/rename_impl.hh @@ -37,6 +37,8 @@ #include "cpu/o3/rename.hh" #include "params/DerivO3CPU.hh" +using namespace std; + template DefaultRename::DefaultRename(O3CPU *_cpu, DerivO3CPUParams *params) : cpu(_cpu), @@ -52,22 +54,22 @@ DefaultRename::DefaultRename(O3CPU *_cpu, DerivO3CPUParams *params) { _status = Inactive; - for (int i=0; i< numThreads; i++) { - renameStatus[i] = Idle; + for (ThreadID tid = 0; tid < numThreads; tid++) { + renameStatus[tid] = Idle; - freeEntries[i].iqEntries = 0; - freeEntries[i].lsqEntries = 0; - freeEntries[i].robEntries = 0; + freeEntries[tid].iqEntries = 0; + freeEntries[tid].lsqEntries = 0; + freeEntries[tid].robEntries = 0; - stalls[i].iew = false; - stalls[i].commit = false; - serializeInst[i] = NULL; + stalls[tid].iew = false; + stalls[tid].commit = false; + serializeInst[tid] = NULL; - instsInProgress[i] = 0; + instsInProgress[tid] = 0; - emptyROB[i] = true; + emptyROB[tid] = true; - serializeOnNextInst[i] = false; + serializeOnNextInst[tid] = false; } // @todo: Make into a parameter. @@ -207,7 +209,7 @@ void DefaultRename::initStage() { // Grab the number of free entries directly from the stages. - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { freeEntries[tid].iqEntries = iew_ptr->instQueue.numFreeEntries(tid); freeEntries[tid].lsqEntries = iew_ptr->ldstQueue.numFreeEntries(tid); freeEntries[tid].robEntries = commit_ptr->numROBFreeEntries(tid); @@ -217,7 +219,7 @@ DefaultRename::initStage() template void -DefaultRename::setActiveThreads(std::list *at_ptr) +DefaultRename::setActiveThreads(list *at_ptr) { activeThreads = at_ptr; } @@ -227,9 +229,8 @@ template void DefaultRename::setRenameMap(RenameMap rm_ptr[]) { - for (int i=0; i @@ -260,19 +261,19 @@ void DefaultRename::switchOut() { // Clear any state, fix up the rename map. - for (int i = 0; i < numThreads; i++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { typename std::list::iterator hb_it = - historyBuffer[i].begin(); + historyBuffer[tid].begin(); - while (!historyBuffer[i].empty()) { - assert(hb_it != historyBuffer[i].end()); + while (!historyBuffer[tid].empty()) { + assert(hb_it != historyBuffer[tid].end()); DPRINTF(Rename, "[tid:%u]: Removing history entry with sequence " - "number %i.\n", i, (*hb_it).instSeqNum); + "number %i.\n", tid, (*hb_it).instSeqNum); // Tell the rename map to set the architected register to the // previous physical register that it was renamed to. - renameMap[i]->setEntry(hb_it->archReg, hb_it->prevPhysReg); + renameMap[tid]->setEntry(hb_it->archReg, hb_it->prevPhysReg); // Put the renamed physical register back on the free list. freeList->addReg(hb_it->newPhysReg); @@ -282,10 +283,10 @@ DefaultRename::switchOut() scoreboard->setReg(hb_it->newPhysReg); } - historyBuffer[i].erase(hb_it++); + historyBuffer[tid].erase(hb_it++); } - insts[i].clear(); - skidBuffer[i].clear(); + insts[tid].clear(); + skidBuffer[tid].clear(); } } @@ -297,24 +298,24 @@ DefaultRename::takeOverFrom() initStage(); // Reset all state prior to taking over from the other CPU. - for (int i=0; i< numThreads; i++) { - renameStatus[i] = Idle; + for (ThreadID tid = 0; tid < numThreads; tid++) { + renameStatus[tid] = Idle; - stalls[i].iew = false; - stalls[i].commit = false; - serializeInst[i] = NULL; + stalls[tid].iew = false; + stalls[tid].commit = false; + serializeInst[tid] = NULL; - instsInProgress[i] = 0; + instsInProgress[tid] = 0; - emptyROB[i] = true; + emptyROB[tid] = true; - serializeOnNextInst[i] = false; + serializeOnNextInst[tid] = false; } } template void -DefaultRename::squash(const InstSeqNum &squash_seq_num, unsigned tid) +DefaultRename::squash(const InstSeqNum &squash_seq_num, ThreadID tid) { DPRINTF(Rename, "[tid:%u]: Squashing instructions.\n",tid); @@ -380,12 +381,12 @@ DefaultRename::tick() sortInsts(); - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); // Check stall and squash signals. while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; DPRINTF(Rename, "Processing [tid:%i]\n", tid); @@ -406,7 +407,7 @@ DefaultRename::tick() threads = activeThreads->begin(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; // If we committed this cycle then doneSeqNum will be > 0 if (fromCommit->commitInfo[tid].doneSeqNum != 0 && @@ -419,7 +420,7 @@ DefaultRename::tick() } // @todo: make into updateProgress function - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { instsInProgress[tid] -= fromIEW->iewInfo[tid].dispatched; assert(instsInProgress[tid] >=0); @@ -429,7 +430,7 @@ DefaultRename::tick() template void -DefaultRename::rename(bool &status_change, unsigned tid) +DefaultRename::rename(bool &status_change, ThreadID tid) { // If status is Running or idle, // call renameInsts() @@ -483,7 +484,7 @@ DefaultRename::rename(bool &status_change, unsigned tid) template void -DefaultRename::renameInsts(unsigned tid) +DefaultRename::renameInsts(ThreadID tid) { // Instructions can be either in the skid buffer or the queue of // instructions coming from decode, depending on the status. @@ -703,7 +704,7 @@ DefaultRename::renameInsts(unsigned tid) template void -DefaultRename::skidInsert(unsigned tid) +DefaultRename::skidInsert(ThreadID tid) { DynInstPtr inst = NULL; @@ -742,8 +743,8 @@ DefaultRename::sortInsts() { int insts_from_decode = fromDecode->size; #ifdef DEBUG - for (int i=0; i < numThreads; i++) - assert(insts[i].empty()); + for (ThreadID tid = 0; tid < numThreads; tid++) + assert(insts[tid].empty()); #endif for (int i = 0; i < insts_from_decode; ++i) { DynInstPtr inst = fromDecode->insts[i]; @@ -755,11 +756,11 @@ template bool DefaultRename::skidsEmpty() { - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!skidBuffer[tid].empty()) return false; @@ -774,11 +775,11 @@ DefaultRename::updateStatus() { bool any_unblocking = false; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (renameStatus[tid] == Unblocking) { any_unblocking = true; @@ -809,7 +810,7 @@ DefaultRename::updateStatus() template bool -DefaultRename::block(unsigned tid) +DefaultRename::block(ThreadID tid) { DPRINTF(Rename, "[tid:%u]: Blocking.\n", tid); @@ -843,7 +844,7 @@ DefaultRename::block(unsigned tid) template bool -DefaultRename::unblock(unsigned tid) +DefaultRename::unblock(ThreadID tid) { DPRINTF(Rename, "[tid:%u]: Trying to unblock.\n", tid); @@ -864,7 +865,7 @@ DefaultRename::unblock(unsigned tid) template void -DefaultRename::doSquash(const InstSeqNum &squashed_seq_num, unsigned tid) +DefaultRename::doSquash(const InstSeqNum &squashed_seq_num, ThreadID tid) { typename std::list::iterator hb_it = historyBuffer[tid].begin(); @@ -904,7 +905,7 @@ DefaultRename::doSquash(const InstSeqNum &squashed_seq_num, unsigned tid) template void -DefaultRename::removeFromHistory(InstSeqNum inst_seq_num, unsigned tid) +DefaultRename::removeFromHistory(InstSeqNum inst_seq_num, ThreadID tid) { DPRINTF(Rename, "[tid:%u]: Removing a committed instruction from the " "history buffer %u (size=%i), until [sn:%lli].\n", @@ -945,7 +946,7 @@ DefaultRename::removeFromHistory(InstSeqNum inst_seq_num, unsigned tid) template inline void -DefaultRename::renameSrcRegs(DynInstPtr &inst,unsigned tid) +DefaultRename::renameSrcRegs(DynInstPtr &inst, ThreadID tid) { assert(renameMap[tid] != 0); @@ -996,7 +997,7 @@ DefaultRename::renameSrcRegs(DynInstPtr &inst,unsigned tid) template inline void -DefaultRename::renameDestRegs(DynInstPtr &inst,unsigned tid) +DefaultRename::renameDestRegs(DynInstPtr &inst, ThreadID tid) { typename RenameMap::RenameInfo rename_result; @@ -1057,7 +1058,7 @@ DefaultRename::renameDestRegs(DynInstPtr &inst,unsigned tid) template inline int -DefaultRename::calcFreeROBEntries(unsigned tid) +DefaultRename::calcFreeROBEntries(ThreadID tid) { int num_free = freeEntries[tid].robEntries - (instsInProgress[tid] - fromIEW->iewInfo[tid].dispatched); @@ -1069,7 +1070,7 @@ DefaultRename::calcFreeROBEntries(unsigned tid) template inline int -DefaultRename::calcFreeIQEntries(unsigned tid) +DefaultRename::calcFreeIQEntries(ThreadID tid) { int num_free = freeEntries[tid].iqEntries - (instsInProgress[tid] - fromIEW->iewInfo[tid].dispatched); @@ -1081,7 +1082,7 @@ DefaultRename::calcFreeIQEntries(unsigned tid) template inline int -DefaultRename::calcFreeLSQEntries(unsigned tid) +DefaultRename::calcFreeLSQEntries(ThreadID tid) { int num_free = freeEntries[tid].lsqEntries - (instsInProgress[tid] - fromIEW->iewInfo[tid].dispatchedToLSQ); @@ -1107,7 +1108,7 @@ DefaultRename::validInsts() template void -DefaultRename::readStallSignals(unsigned tid) +DefaultRename::readStallSignals(ThreadID tid) { if (fromIEW->iewBlock[tid]) { stalls[tid].iew = true; @@ -1130,7 +1131,7 @@ DefaultRename::readStallSignals(unsigned tid) template bool -DefaultRename::checkStall(unsigned tid) +DefaultRename::checkStall(ThreadID tid) { bool ret_val = false; @@ -1165,7 +1166,7 @@ DefaultRename::checkStall(unsigned tid) template void -DefaultRename::readFreeEntries(unsigned tid) +DefaultRename::readFreeEntries(ThreadID tid) { bool updated = false; if (fromIEW->iewInfo[tid].usedIQ) { @@ -1199,7 +1200,7 @@ DefaultRename::readFreeEntries(unsigned tid) template bool -DefaultRename::checkSignalsAndUpdate(unsigned tid) +DefaultRename::checkSignalsAndUpdate(ThreadID tid) { // Check if there's a squash signal, squash if there is // Check stall signals, block if necessary. @@ -1308,8 +1309,7 @@ DefaultRename::checkSignalsAndUpdate(unsigned tid) template void -DefaultRename::serializeAfter(InstQueue &inst_list, - unsigned tid) +DefaultRename::serializeAfter(InstQueue &inst_list, ThreadID tid) { if (inst_list.empty()) { // Mark a bit to say that I must serialize on the next instruction. @@ -1347,11 +1347,11 @@ DefaultRename::dumpHistory() { typename std::list::iterator buf_it; - for (int i = 0; i < numThreads; i++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { - buf_it = historyBuffer[i].begin(); + buf_it = historyBuffer[tid].begin(); - while (buf_it != historyBuffer[i].end()) { + while (buf_it != historyBuffer[tid].end()) { cprintf("Seq num: %i\nArch reg: %i New phys reg: %i Old phys " "reg: %i\n", (*buf_it).instSeqNum, (int)(*buf_it).archReg, (int)(*buf_it).newPhysReg, (int)(*buf_it).prevPhysReg); diff --git a/src/cpu/o3/rob.hh b/src/cpu/o3/rob.hh index 00329abb0..657bc8d06 100644 --- a/src/cpu/o3/rob.hh +++ b/src/cpu/o3/rob.hh @@ -84,14 +84,14 @@ class ROB */ ROB(O3CPU *_cpu, unsigned _numEntries, unsigned _squashWidth, std::string smtROBPolicy, unsigned _smtROBThreshold, - unsigned _numThreads); + ThreadID _numThreads); std::string name() const; /** Sets pointer to the list of active threads. * @param at_ptr Pointer to the list of active threads. */ - void setActiveThreads(std::list* at_ptr); + void setActiveThreads(std::list *at_ptr); /** Switches out the ROB. */ void switchOut(); @@ -116,7 +116,7 @@ class ROB * the ROB. * @return Pointer to the DynInst that is at the head of the ROB. */ - DynInstPtr readHeadInst(unsigned tid); + DynInstPtr readHeadInst(ThreadID tid); /** Returns pointer to the tail instruction within the ROB. There is * no guarantee as to the return value if the ROB is empty. @@ -128,7 +128,7 @@ class ROB * the ROB. * @return Pointer to the DynInst that is at the tail of the ROB. */ - DynInstPtr readTailInst(unsigned tid); + DynInstPtr readTailInst(ThreadID tid); /** Retires the head instruction, removing it from the ROB. */ // void retireHead(); @@ -136,13 +136,13 @@ class ROB /** Retires the head instruction of a specific thread, removing it from the * ROB. */ - void retireHead(unsigned tid); + void retireHead(ThreadID tid); /** Is the oldest instruction across all threads ready. */ // bool isHeadReady(); /** Is the oldest instruction across a particular thread ready. */ - bool isHeadReady(unsigned tid); + bool isHeadReady(ThreadID tid); /** Is there any commitable head instruction across all threads ready. */ bool canCommit(); @@ -151,20 +151,20 @@ class ROB void resetEntries(); /** Number of entries needed For 'num_threads' amount of threads. */ - int entryAmount(int num_threads); + int entryAmount(ThreadID num_threads); /** Returns the number of total free entries in the ROB. */ unsigned numFreeEntries(); /** Returns the number of free entries in a specific ROB paritition. */ - unsigned numFreeEntries(unsigned tid); + unsigned numFreeEntries(ThreadID tid); /** Returns the maximum number of entries for a specific thread. */ - unsigned getMaxEntries(unsigned tid) + unsigned getMaxEntries(ThreadID tid) { return maxEntries[tid]; } /** Returns the number of entries being used by a specific thread. */ - unsigned getThreadEntries(unsigned tid) + unsigned getThreadEntries(ThreadID tid) { return threadEntries[tid]; } /** Returns if the ROB is full. */ @@ -172,7 +172,7 @@ class ROB { return numInstsInROB == numEntries; } /** Returns if a specific thread's partition is full. */ - bool isFull(unsigned tid) + bool isFull(ThreadID tid) { return threadEntries[tid] == numEntries; } /** Returns if the ROB is empty. */ @@ -180,16 +180,16 @@ class ROB { return numInstsInROB == 0; } /** Returns if a specific thread's partition is empty. */ - bool isEmpty(unsigned tid) + bool isEmpty(ThreadID tid) { return threadEntries[tid] == 0; } /** Executes the squash, marking squashed instructions. */ - void doSquash(unsigned tid); + void doSquash(ThreadID tid); /** Squashes all instructions younger than the given sequence number for * the specific thread. */ - void squash(InstSeqNum squash_num, unsigned tid); + void squash(InstSeqNum squash_num, ThreadID tid); /** Updates the head instruction with the new oldest instruction. */ void updateHead(); @@ -201,37 +201,37 @@ class ROB // uint64_t readHeadPC(); /** Reads the PC of the head instruction of a specific thread. */ -// uint64_t readHeadPC(unsigned tid); +// uint64_t readHeadPC(ThreadID tid); /** Reads the next PC of the oldest head instruction. */ // uint64_t readHeadNextPC(); /** Reads the next PC of the head instruction of a specific thread. */ -// uint64_t readHeadNextPC(unsigned tid); +// uint64_t readHeadNextPC(ThreadID tid); /** Reads the sequence number of the oldest head instruction. */ // InstSeqNum readHeadSeqNum(); /** Reads the sequence number of the head instruction of a specific thread. */ -// InstSeqNum readHeadSeqNum(unsigned tid); +// InstSeqNum readHeadSeqNum(ThreadID tid); /** Reads the PC of the youngest tail instruction. */ // uint64_t readTailPC(); /** Reads the PC of the tail instruction of a specific thread. */ -// uint64_t readTailPC(unsigned tid); +// uint64_t readTailPC(ThreadID tid); /** Reads the sequence number of the youngest tail instruction. */ // InstSeqNum readTailSeqNum(); /** Reads the sequence number of tail instruction of a specific thread. */ -// InstSeqNum readTailSeqNum(unsigned tid); +// InstSeqNum readTailSeqNum(ThreadID tid); /** Checks if the ROB is still in the process of squashing instructions. * @retval Whether or not the ROB is done squashing. */ - bool isDoneSquashing(unsigned tid) const + bool isDoneSquashing(ThreadID tid) const { return doneSquashing[tid]; } /** Checks if the ROB is still in the process of squashing instructions for @@ -249,14 +249,14 @@ class ROB * threadEntries to get the instructions in the ROB unless you are * double checking that variable. */ - int countInsts(unsigned tid); + int countInsts(ThreadID tid); private: /** Pointer to the CPU. */ O3CPU *cpu; /** Active Threads in CPU */ - std::list* activeThreads; + std::list *activeThreads; /** Number of instructions in the ROB. */ unsigned numEntries; @@ -309,7 +309,7 @@ class ROB bool doneSquashing[Impl::MaxThreads]; /** Number of active threads. */ - unsigned numThreads; + ThreadID numThreads; }; #endif //__CPU_O3_ROB_HH__ diff --git a/src/cpu/o3/rob_impl.hh b/src/cpu/o3/rob_impl.hh index 7ff3aa274..8acaa8ecb 100644 --- a/src/cpu/o3/rob_impl.hh +++ b/src/cpu/o3/rob_impl.hh @@ -29,22 +29,24 @@ * Korey Sewell */ +#include + #include "config/full_system.hh" #include "cpu/o3/rob.hh" -#include +using namespace std; template ROB::ROB(O3CPU *_cpu, unsigned _numEntries, unsigned _squashWidth, std::string _smtROBPolicy, unsigned _smtROBThreshold, - unsigned _numThreads) + ThreadID _numThreads) : cpu(_cpu), numEntries(_numEntries), squashWidth(_squashWidth), numInstsInROB(0), numThreads(_numThreads) { - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { squashedSeqNum[tid] = 0; doneSquashing[tid] = true; threadEntries[tid] = 0; @@ -61,8 +63,8 @@ ROB::ROB(O3CPU *_cpu, unsigned _numEntries, unsigned _squashWidth, robPolicy = Dynamic; //Set Max Entries to Total ROB Capacity - for (int i = 0; i < numThreads; i++) { - maxEntries[i]=numEntries; + for (ThreadID tid = 0; tid < numThreads; tid++) { + maxEntries[tid] = numEntries; } } else if (policy == "partitioned") { @@ -73,8 +75,8 @@ ROB::ROB(O3CPU *_cpu, unsigned _numEntries, unsigned _squashWidth, int part_amt = numEntries / numThreads; //Divide ROB up evenly - for (int i = 0; i < numThreads; i++) { - maxEntries[i]=part_amt; + for (ThreadID tid = 0; tid < numThreads; tid++) { + maxEntries[tid] = part_amt; } } else if (policy == "threshold") { @@ -84,8 +86,8 @@ ROB::ROB(O3CPU *_cpu, unsigned _numEntries, unsigned _squashWidth, int threshold = _smtROBThreshold;; //Divide up by threshold amount - for (int i = 0; i < numThreads; i++) { - maxEntries[i]=threshold; + for (ThreadID tid = 0; tid < numThreads; tid++) { + maxEntries[tid] = threshold; } } else { assert(0 && "Invalid ROB Sharing Policy.Options Are:{Dynamic," @@ -93,8 +95,8 @@ ROB::ROB(O3CPU *_cpu, unsigned _numEntries, unsigned _squashWidth, } // Set the per-thread iterators to the end of the instruction list. - for (int i=0; i < numThreads;i++) { - squashIt[i] = instList[i].end(); + for (ThreadID tid = 0; tid < numThreads; tid++) { + squashIt[tid] = instList[tid].end(); } // Initialize the "universal" ROB head & tail point to invalid @@ -112,7 +114,7 @@ ROB::name() const template void -ROB::setActiveThreads(std::list *at_ptr) +ROB::setActiveThreads(list *at_ptr) { DPRINTF(ROB, "Setting active threads list pointer.\n"); activeThreads = at_ptr; @@ -122,7 +124,7 @@ template void ROB::switchOut() { - for (int tid = 0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { instList[tid].clear(); } } @@ -131,7 +133,7 @@ template void ROB::takeOverFrom() { - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { doneSquashing[tid] = true; threadEntries[tid] = 0; squashIt[tid] = instList[tid].end(); @@ -151,11 +153,11 @@ ROB::resetEntries() if (robPolicy != Dynamic || numThreads > 1) { int active_threads = activeThreads->size(); - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (robPolicy == Partitioned) { maxEntries[tid] = numEntries / active_threads; @@ -168,7 +170,7 @@ ROB::resetEntries() template int -ROB::entryAmount(int num_threads) +ROB::entryAmount(ThreadID num_threads) { if (robPolicy == Partitioned) { return numEntries / num_threads; @@ -181,17 +183,17 @@ template int ROB::countInsts() { - int total=0; + int total = 0; - for (int i=0;i < numThreads;i++) - total += countInsts(i); + for (ThreadID tid = 0; tid < numThreads; tid++) + total += countInsts(tid); return total; } template int -ROB::countInsts(unsigned tid) +ROB::countInsts(ThreadID tid) { return instList[tid].size(); } @@ -207,7 +209,7 @@ ROB::insertInst(DynInstPtr &inst) assert(numInstsInROB != numEntries); - int tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; instList[tid].push_back(inst); @@ -242,7 +244,7 @@ ROB::retireHead() //assert(numInstsInROB == countInsts()); assert(numInstsInROB > 0); - int tid = (*head)->threadNumber; + ThreadID tid = (*head)->threadNumber; retireHead(tid); @@ -254,7 +256,7 @@ ROB::retireHead() template void -ROB::retireHead(unsigned tid) +ROB::retireHead(ThreadID tid) { //assert(numInstsInROB == countInsts()); assert(numInstsInROB > 0); @@ -300,7 +302,7 @@ ROB::isHeadReady() */ template bool -ROB::isHeadReady(unsigned tid) +ROB::isHeadReady(ThreadID tid) { if (threadEntries[tid] != 0) { return instList[tid].front()->readyToCommit(); @@ -314,11 +316,11 @@ bool ROB::canCommit() { //@todo: set ActiveThreads through ROB or CPU - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (isHeadReady(tid)) { return true; @@ -339,14 +341,14 @@ ROB::numFreeEntries() template unsigned -ROB::numFreeEntries(unsigned tid) +ROB::numFreeEntries(ThreadID tid) { return maxEntries[tid] - threadEntries[tid]; } template void -ROB::doSquash(unsigned tid) +ROB::doSquash(ThreadID tid) { DPRINTF(ROB, "[tid:%u]: Squashing instructions until [sn:%i].\n", tid, squashedSeqNum[tid]); @@ -429,11 +431,11 @@ ROB::updateHead() bool first_valid = true; // @todo: set ActiveThreads through ROB or CPU - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (instList[tid].empty()) continue; @@ -470,11 +472,11 @@ ROB::updateTail() tail = instList[0].end(); bool first_valid = true; - std::list::iterator threads = activeThreads->begin(); - std::list::iterator end = activeThreads->end(); + list::iterator threads = activeThreads->begin(); + list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (instList[tid].empty()) { continue; @@ -503,7 +505,7 @@ ROB::updateTail() template void -ROB::squash(InstSeqNum squash_num,unsigned tid) +ROB::squash(InstSeqNum squash_num, ThreadID tid) { if (isEmpty()) { DPRINTF(ROB, "Does not need to squash due to being empty " @@ -546,7 +548,7 @@ ROB::readHeadInst() template typename Impl::DynInstPtr -ROB::readHeadInst(unsigned tid) +ROB::readHeadInst(ThreadID tid) { if (threadEntries[tid] != 0) { InstIt head_thread = instList[tid].begin(); @@ -573,7 +575,7 @@ ROB::readHeadPC() template uint64_t -ROB::readHeadPC(unsigned tid) +ROB::readHeadPC(ThreadID tid) { //assert(numInstsInROB == countInsts()); InstIt head_thread = instList[tid].begin(); @@ -595,7 +597,7 @@ ROB::readHeadNextPC() template uint64_t -ROB::readHeadNextPC(unsigned tid) +ROB::readHeadNextPC(ThreadID tid) { //assert(numInstsInROB == countInsts()); InstIt head_thread = instList[tid].begin(); @@ -615,7 +617,7 @@ ROB::readHeadSeqNum() template InstSeqNum -ROB::readHeadSeqNum(unsigned tid) +ROB::readHeadSeqNum(ThreadID tid) { InstIt head_thread = instList[tid].begin(); @@ -634,7 +636,7 @@ ROB::readTailInst() */ template typename Impl::DynInstPtr -ROB::readTailInst(unsigned tid) +ROB::readTailInst(ThreadID tid) { //assert(tail_thread[tid] != instList[tid].end()); @@ -658,7 +660,7 @@ ROB::readTailPC() template uint64_t -ROB::readTailPC(unsigned tid) +ROB::readTailPC(ThreadID tid) { //assert(tail_thread[tid] != instList[tid].end()); @@ -680,7 +682,7 @@ ROB::readTailSeqNum() template InstSeqNum -ROB::readTailSeqNum(unsigned tid) +ROB::readTailSeqNum(ThreadID tid) { // Return the last sequence number that has not been squashed. Other // stages can use it to squash any instructions younger than the current diff --git a/src/cpu/o3/store_set.cc b/src/cpu/o3/store_set.cc index 2d28b617f..df4ee00ad 100644 --- a/src/cpu/o3/store_set.cc +++ b/src/cpu/o3/store_set.cc @@ -186,8 +186,7 @@ StoreSet::insertLoad(Addr load_PC, InstSeqNum load_seq_num) } void -StoreSet::insertStore(Addr store_PC, InstSeqNum store_seq_num, - unsigned tid) +StoreSet::insertStore(Addr store_PC, InstSeqNum store_seq_num, ThreadID tid) { int index = calcIndex(store_PC); @@ -288,7 +287,7 @@ StoreSet::issued(Addr issued_PC, InstSeqNum issued_seq_num, bool is_store) } void -StoreSet::squash(InstSeqNum squashed_num, unsigned tid) +StoreSet::squash(InstSeqNum squashed_num, ThreadID tid) { DPRINTF(StoreSet, "StoreSet: Squashing until inum %i\n", squashed_num); diff --git a/src/cpu/o3/store_set.hh b/src/cpu/o3/store_set.hh index 57cd2a197..ce4591f68 100644 --- a/src/cpu/o3/store_set.hh +++ b/src/cpu/o3/store_set.hh @@ -82,8 +82,7 @@ class StoreSet /** Inserts a store into the store set predictor. Updates the * LFST if the store has a valid SSID. */ - void insertStore(Addr store_PC, InstSeqNum store_seq_num, - unsigned tid); + void insertStore(Addr store_PC, InstSeqNum store_seq_num, ThreadID tid); /** Checks if the instruction with the given PC is dependent upon * any store. @return Returns the sequence number of the store @@ -95,7 +94,7 @@ class StoreSet void issued(Addr issued_PC, InstSeqNum issued_seq_num, bool is_store); /** Squashes for a specific thread until the given sequence number. */ - void squash(InstSeqNum squashed_num, unsigned tid); + void squash(InstSeqNum squashed_num, ThreadID tid); /** Resets all tables. */ void clear(); diff --git a/src/cpu/o3/thread_context_impl.hh b/src/cpu/o3/thread_context_impl.hh index 07140a19f..bce334dc4 100755 --- a/src/cpu/o3/thread_context_impl.hh +++ b/src/cpu/o3/thread_context_impl.hh @@ -226,7 +226,7 @@ O3ThreadContext::copyArchRegs(ThreadContext *tc) { // This function will mess things up unless the ROB is empty and // there are no instructions in the pipeline. - unsigned tid = thread->threadId(); + ThreadID tid = thread->threadId(); PhysRegIndex renamed_reg; // First loop through the integer registers. diff --git a/src/cpu/ozone/back_end_impl.hh b/src/cpu/ozone/back_end_impl.hh index 7753a83b5..fbb363a96 100644 --- a/src/cpu/ozone/back_end_impl.hh +++ b/src/cpu/ozone/back_end_impl.hh @@ -649,28 +649,28 @@ BackEnd::regStats() { using namespace Stats; rob_cap_events - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ROB:cap_events") .desc("number of cycles where ROB cap was active") .flags(total) ; rob_cap_inst_count - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ROB:cap_inst") .desc("number of instructions held up by ROB cap") .flags(total) ; iq_cap_events - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() +".IQ:cap_events" ) .desc("number of cycles where IQ cap was active") .flags(total) ; iq_cap_inst_count - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".IQ:cap_inst") .desc("number of instructions held up by IQ cap") .flags(total) @@ -678,49 +678,49 @@ BackEnd::regStats() exe_inst - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:count") .desc("number of insts issued") .flags(total) ; exe_swp - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:swp") .desc("number of swp insts issued") .flags(total) ; exe_nop - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:nop") .desc("number of nop insts issued") .flags(total) ; exe_refs - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:refs") .desc("number of memory reference insts issued") .flags(total) ; exe_loads - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:loads") .desc("number of load insts issued") .flags(total) ; exe_branches - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:branches") .desc("Number of branches issued") .flags(total) ; issued_ops - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:op_count") .desc("number of insts issued") .flags(total) @@ -737,28 +737,28 @@ BackEnd::regStats() // Other stats // lsq_forw_loads - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".LSQ:forw_loads") .desc("number of loads forwarded via LSQ") .flags(total) ; inv_addr_loads - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:addr_loads") .desc("number of invalid-address loads") .flags(total) ; inv_addr_swpfs - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:addr_swpfs") .desc("number of invalid-address SW prefetches") .flags(total) ; lsq_blocked_loads - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".LSQ:blocked_loads") .desc("number of ready loads not issued due to memory disambiguation") .flags(total) @@ -793,28 +793,28 @@ BackEnd::regStats() } writeback_count - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:count") .desc("cumulative count of insts written-back") .flags(total) ; producer_inst - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:producers") .desc("num instructions producing a value") .flags(total) ; consumer_inst - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:consumers") .desc("num instructions consuming a value") .flags(total) ; wb_penalized - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:penalized") .desc("number of instrctions required to write to 'other' IQ") .flags(total) @@ -845,42 +845,42 @@ BackEnd::regStats() wb_rate = writeback_count / cpu->numCycles; stat_com_inst - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:count") .desc("Number of instructions committed") .flags(total) ; stat_com_swp - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:swp_count") .desc("Number of s/w prefetches committed") .flags(total) ; stat_com_refs - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:refs") .desc("Number of memory references committed") .flags(total) ; stat_com_loads - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:loads") .desc("Number of loads committed") .flags(total) ; stat_com_membars - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:membars") .desc("Number of memory barriers committed") .flags(total) ; stat_com_branches - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:branches") .desc("Number of branches committed") .flags(total) @@ -906,7 +906,7 @@ BackEnd::regStats() // we reached the BW limit // commit_eligible - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:bw_limited") .desc("number of insts not committed due to BW limits") .flags(total) @@ -923,7 +923,7 @@ BackEnd::regStats() ; ROB_count - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ROB:occupancy") .desc(name() + ".ROB occupancy (cumulative)") .flags(total) @@ -943,7 +943,7 @@ BackEnd::regStats() ROB_occ_rate = ROB_count / cpu->numCycles; ROB_occ_dist - .init(cpu->number_of_threads,0,numROBEntries,2) + .init(cpu->numThreads, 0, numROBEntries, 2) .name(name() + ".ROB:occ_dist") .desc("ROB Occupancy per cycle") .flags(total | cdf) @@ -1292,7 +1292,7 @@ BackEnd::executeInsts() // This probably needs to prioritize the redirects if a different // scheduler is used. Currently the scheduler schedules the oldest // instruction first, so the branch resolution order will be correct. - unsigned tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; if (!fetchRedirect[tid]) { @@ -1735,34 +1735,34 @@ template void BackEnd::updateExeInstStats(DynInstPtr &inst) { - int thread_number = inst->threadNumber; + ThreadID tid = inst->threadNumber; // // Pick off the software prefetches // #ifdef TARGET_ALPHA if (inst->isDataPrefetch()) - exe_swp[thread_number]++; + exe_swp[tid]++; else - exe_inst[thread_number]++; + exe_inst[tid]++; #else - exe_inst[thread_number]++; + exe_inst[tid]++; #endif // // Control operations // if (inst->isControl()) - exe_branches[thread_number]++; + exe_branches[tid]++; // // Memory operations // if (inst->isMemRef()) { - exe_refs[thread_number]++; + exe_refs[tid]++; if (inst->isLoad()) - exe_loads[thread_number]++; + exe_loads[tid]++; } } @@ -1770,40 +1770,40 @@ template void BackEnd::updateComInstStats(DynInstPtr &inst) { - unsigned thread = inst->threadNumber; + ThreadID tid = inst->threadNumber; // // Pick off the software prefetches // #ifdef TARGET_ALPHA if (inst->isDataPrefetch()) { - stat_com_swp[thread]++; + stat_com_swp[tid]++; } else { - stat_com_inst[thread]++; + stat_com_inst[tid]++; } #else - stat_com_inst[thread]++; + stat_com_inst[tid]++; #endif // // Control Instructions // if (inst->isControl()) - stat_com_branches[thread]++; + stat_com_branches[tid]++; // // Memory references // if (inst->isMemRef()) { - stat_com_refs[thread]++; + stat_com_refs[tid]++; if (inst->isLoad()) { - stat_com_loads[thread]++; + stat_com_loads[tid]++; } } if (inst->isMemBarrier()) { - stat_com_membars[thread]++; + stat_com_membars[tid]++; } } diff --git a/src/cpu/ozone/cpu_builder.cc b/src/cpu/ozone/cpu_builder.cc index beb179d41..9cac3a524 100644 --- a/src/cpu/ozone/cpu_builder.cc +++ b/src/cpu/ozone/cpu_builder.cc @@ -59,11 +59,11 @@ DerivOzoneCPUParams::create() #if FULL_SYSTEM // Full-system only supports a single thread for the moment. - int actual_num_threads = 1; + ThreadID actual_num_threads = 1; #else // In non-full-system mode, we infer the number of threads from // the workload if it's not explicitly specified. - int actual_num_threads = + ThreadID actual_num_threads = numThreads.isValid() ? numThreads : workload.size(); if (workload.size() == 0) { diff --git a/src/cpu/ozone/inst_queue.hh b/src/cpu/ozone/inst_queue.hh index 5af916fb5..2f544e307 100644 --- a/src/cpu/ozone/inst_queue.hh +++ b/src/cpu/ozone/inst_queue.hh @@ -130,7 +130,7 @@ class InstQueue void setTimeBuffer(TimeBuffer *tb_ptr); /** Number of entries needed for given amount of threads. */ - int entryAmount(int num_threads); + int entryAmount(ThreadID num_threads); /** Resets max entries for all threads. */ void resetEntries(); @@ -139,13 +139,13 @@ class InstQueue unsigned numFreeEntries(); /** Returns number of free entries for a thread. */ - unsigned numFreeEntries(unsigned tid); + unsigned numFreeEntries(ThreadID tid); /** Returns whether or not the IQ is full. */ bool isFull(); /** Returns whether or not the IQ is full for a specific thread. */ - bool isFull(unsigned tid); + bool isFull(ThreadID tid); /** Returns if there are any ready instructions in the IQ. */ bool hasReadyInsts(); @@ -179,7 +179,7 @@ class InstQueue * Commits all instructions up to and including the given sequence number, * for a specific thread. */ - void commit(const InstSeqNum &inst, unsigned tid = 0); + void commit(const InstSeqNum &inst, ThreadID tid = 0); /** Wakes all dependents of a completed instruction. */ void wakeDependents(DynInstPtr &completed_inst); @@ -206,10 +206,10 @@ class InstQueue * Squashes instructions for a thread. Squashing information is obtained * from the time buffer. */ - void squash(unsigned tid); // Probably want the ISN + void squash(ThreadID tid); // Probably want the ISN /** Returns the number of used entries for a thread. */ - unsigned getCount(unsigned tid) { return count[tid]; }; + unsigned getCount(ThreadID tid) { return count[tid]; }; /** Updates the number of free entries. */ void updateFreeEntries(int num) { freeEntries += num; } @@ -219,7 +219,7 @@ class InstQueue private: /** Does the actual squashing. */ - void doSquash(unsigned tid); + void doSquash(ThreadID tid); ///////////////////////// // Various pointers diff --git a/src/cpu/ozone/inst_queue_impl.hh b/src/cpu/ozone/inst_queue_impl.hh index 3c3084757..ae2e3b09b 100644 --- a/src/cpu/ozone/inst_queue_impl.hh +++ b/src/cpu/ozone/inst_queue_impl.hh @@ -83,8 +83,8 @@ InstQueue::InstQueue(Params *params) numThreads = 1; //Initialize thread IQ counts - for (int i = 0; i ::InstQueue(Params *params) // regScoreboard.resize(numPhysRegs); /* //Initialize Mem Dependence Units - for (int i = 0; i < numThreads; i++) { - memDepUnit[i].init(params,i); - memDepUnit[i].setIQ(this); + for (ThreadID tid = 0; tid < numThreads; tid++) { + memDepUnit[tid].init(params, tid); + memDepUnit[tid].setIQ(this); } // Initialize all the head pointers to point to NULL, and all the @@ -121,8 +121,8 @@ InstQueue::InstQueue(Params *params) regScoreboard[i] = false; } */ - for (int i = 0; i < numThreads; ++i) { - squashedSeqNum[i] = 0; + for (ThreadID tid = 0; tid < numThreads; ++tid) { + squashedSeqNum[tid] = 0; } /* for (int i = 0; i < Num_OpClasses; ++i) { @@ -141,8 +141,8 @@ InstQueue::InstQueue(Params *params) iqPolicy = Dynamic; //Set Max Entries to Total ROB Capacity - for (int i = 0; i < numThreads; i++) { - maxEntries[i] = numEntries; + for (ThreadID tid = 0; tid < numThreads; tid++) { + maxEntries[tid] = numEntries; } } else if (policy == "partitioned") { @@ -152,8 +152,8 @@ InstQueue::InstQueue(Params *params) int part_amt = numEntries / numThreads; //Divide ROB up evenly - for (int i = 0; i < numThreads; i++) { - maxEntries[i] = part_amt; + for (ThreadID tid = 0; tid < numThreads; tid++) { + maxEntries[tid] = part_amt; } DPRINTF(Fetch, "IQ sharing policy set to Partitioned:" @@ -167,8 +167,8 @@ InstQueue::InstQueue(Params *params) int thresholdIQ = (int)((double)threshold * numEntries); //Divide up by threshold amount - for (int i = 0; i < numThreads; i++) { - maxEntries[i] = thresholdIQ; + for (ThreadID tid = 0; tid < numThreads; tid++) { + maxEntries[tid] = thresholdIQ; } DPRINTF(Fetch, "IQ sharing policy set to Threshold:" @@ -292,9 +292,9 @@ InstQueue::regStats() .desc("Number of squashed non-spec instructions that were removed") .prereq(iqSquashedNonSpecRemoved); /* - for ( int i=0; i < numThreads; i++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { // Tell mem dependence unit to reg stats as well. - memDepUnit[i].regStats(); + memDepUnit[tid].regStats(); } */ } @@ -327,7 +327,7 @@ InstQueue::setTimeBuffer(TimeBuffer *tb_ptr) template int -InstQueue::entryAmount(int num_threads) +InstQueue::entryAmount(ThreadID num_threads) { if (iqPolicy == Partitioned) { return numEntries / num_threads; @@ -348,7 +348,7 @@ InstQueue::resetEntries() std::list::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (iqPolicy == Partitioned) { maxEntries[tid] = numEntries / active_threads; @@ -368,7 +368,7 @@ InstQueue::numFreeEntries() template unsigned -InstQueue::numFreeEntries(unsigned tid) +InstQueue::numFreeEntries(ThreadID tid) { return maxEntries[tid] - count[tid]; } @@ -388,7 +388,7 @@ InstQueue::isFull() template bool -InstQueue::isFull(unsigned tid) +InstQueue::isFull(ThreadID tid) { if (numFreeEntries(tid) == 0) { return(true); @@ -735,7 +735,7 @@ InstQueue::scheduleNonSpec(const InstSeqNum &inst) assert(inst_it != nonSpecInsts.end()); -// unsigned tid = (*inst_it).second->threadNumber; +// ThreadID tid = (*inst_it).second->threadNumber; // Mark this instruction as ready to issue. (*inst_it).second->setCanIssue(); @@ -752,7 +752,7 @@ InstQueue::scheduleNonSpec(const InstSeqNum &inst) template void -InstQueue::commit(const InstSeqNum &inst, unsigned tid) +InstQueue::commit(const InstSeqNum &inst, ThreadID tid) { /*Need to go through each thread??*/ DPRINTF(IQ, "[tid:%i]: Committing instructions older than [sn:%i]\n", @@ -879,7 +879,7 @@ template void InstQueue::completeMemInst(DynInstPtr &completed_inst) { - int tid = completed_inst->threadNumber; + ThreadID tid = completed_inst->threadNumber; DPRINTF(IQ, "Completing mem instruction PC:%#x [sn:%lli]\n", completed_inst->readPC(), completed_inst->seqNum); @@ -903,7 +903,7 @@ InstQueue::violation(DynInstPtr &store, */ template void -InstQueue::squash(unsigned tid) +InstQueue::squash(ThreadID tid) { DPRINTF(IQ, "[tid:%i]: Starting to squash instructions in " "the IQ.\n", tid); @@ -927,7 +927,7 @@ InstQueue::squash(unsigned tid) template void -InstQueue::doSquash(unsigned tid) +InstQueue::doSquash(ThreadID tid) { // Make sure the squashed sequence number is valid. assert(squashedSeqNum[tid] != 0); @@ -1200,10 +1200,10 @@ InstQueue::countInsts() #if 0 int total_insts = 0; - for (int i = 0; i < numThreads; ++i) { - ListIt count_it = instList[i].begin(); + for (ThreadID tid = 0; tid < numThreads; ++tid) { + ListIt count_it = instList[tid].begin(); - while (count_it != instList[i].end()) { + while (count_it != instList[tid].end()) { if (!(*count_it)->isSquashed() && !(*count_it)->isSquashedInIQ()) { if (!(*count_it)->isIssued()) { ++total_insts; @@ -1300,13 +1300,13 @@ template void InstQueue::dumpInsts() { - for (int i = 0; i < numThreads; ++i) { + for (ThreadID tid = 0; tid < numThreads; ++tid) { // int num = 0; // int valid_num = 0; /* - ListIt inst_list_it = instList[i].begin(); + ListIt inst_list_it = instList[tid].begin(); - while (inst_list_it != instList[i].end()) + while (inst_list_it != instList[tid].end()) { cprintf("Instruction:%i\n", num); diff --git a/src/cpu/ozone/lw_back_end_impl.hh b/src/cpu/ozone/lw_back_end_impl.hh index 60c42edd3..86d4531a0 100644 --- a/src/cpu/ozone/lw_back_end_impl.hh +++ b/src/cpu/ozone/lw_back_end_impl.hh @@ -188,77 +188,77 @@ LWBackEnd::regStats() LSQ.regStats(); robCapEvents - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ROB:cap_events") .desc("number of cycles where ROB cap was active") .flags(total) ; robCapInstCount - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ROB:cap_inst") .desc("number of instructions held up by ROB cap") .flags(total) ; iqCapEvents - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() +".IQ:cap_events" ) .desc("number of cycles where IQ cap was active") .flags(total) ; iqCapInstCount - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".IQ:cap_inst") .desc("number of instructions held up by IQ cap") .flags(total) ; exeInst - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:count") .desc("number of insts issued") .flags(total) ; exeSwp - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:swp") .desc("number of swp insts issued") .flags(total) ; exeNop - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:nop") .desc("number of nop insts issued") .flags(total) ; exeRefs - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:refs") .desc("number of memory reference insts issued") .flags(total) ; exeLoads - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:loads") .desc("number of load insts issued") .flags(total) ; exeBranches - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:branches") .desc("Number of branches issued") .flags(total) ; issuedOps - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:op_count") .desc("number of insts issued") .flags(total) @@ -275,28 +275,28 @@ LWBackEnd::regStats() // Other stats // lsqForwLoads - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".LSQ:forw_loads") .desc("number of loads forwarded via LSQ") .flags(total) ; invAddrLoads - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:addr_loads") .desc("number of invalid-address loads") .flags(total) ; invAddrSwpfs - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ISSUE:addr_swpfs") .desc("number of invalid-address SW prefetches") .flags(total) ; lsqBlockedLoads - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".LSQ:blocked_loads") .desc("number of ready loads not issued due to memory disambiguation") .flags(total) @@ -332,28 +332,28 @@ LWBackEnd::regStats() } */ writebackCount - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:count") .desc("cumulative count of insts written-back") .flags(total) ; producerInst - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:producers") .desc("num instructions producing a value") .flags(total) ; consumerInst - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:consumers") .desc("num instructions consuming a value") .flags(total) ; wbPenalized - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:penalized") .desc("number of instrctions required to write to 'other' IQ") .flags(total) @@ -384,42 +384,42 @@ LWBackEnd::regStats() wbRate = writebackCount / cpu->numCycles; statComInst - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:count") .desc("Number of instructions committed") .flags(total) ; statComSwp - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:swp_count") .desc("Number of s/w prefetches committed") .flags(total) ; statComRefs - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:refs") .desc("Number of memory references committed") .flags(total) ; statComLoads - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:loads") .desc("Number of loads committed") .flags(total) ; statComMembars - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:membars") .desc("Number of memory barriers committed") .flags(total) ; statComBranches - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:branches") .desc("Number of branches committed") .flags(total) @@ -445,7 +445,7 @@ LWBackEnd::regStats() // we reached the BW limit // commitEligible - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:bw_limited") .desc("number of insts not committed due to BW limits") .flags(total) @@ -457,13 +457,13 @@ LWBackEnd::regStats() ; squashedInsts - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:squashed_insts") .desc("Number of instructions removed from inst list") ; ROBSquashedInsts - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:rob_squashed_insts") .desc("Number of instructions removed from inst list when they reached the head of the ROB") ; @@ -474,7 +474,7 @@ LWBackEnd::regStats() ; ROBCount - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".ROB:occupancy") .desc(name() + ".ROB occupancy (cumulative)") .flags(total) @@ -494,7 +494,7 @@ LWBackEnd::regStats() ROBOccRate = ROBCount / cpu->numCycles; /* ROBOccDist - .init(cpu->number_of_threads,0,numROBEntries,2) + .init(cpu->numThreads, 0, numROBEntries, 2) .name(name() + ".ROB:occ_dist") .desc("ROB Occupancy per cycle") .flags(total | cdf) @@ -1498,34 +1498,34 @@ template void LWBackEnd::updateExeInstStats(DynInstPtr &inst) { - int thread_number = inst->threadNumber; + ThreadID tid = inst->threadNumber; // // Pick off the software prefetches // #ifdef TARGET_ALPHA if (inst->isDataPrefetch()) - exeSwp[thread_number]++; + exeSwp[tid]++; else - exeInst[thread_number]++; + exeInst[tid]++; #else - exeInst[thread_number]++; + exeInst[tid]++; #endif // // Control operations // if (inst->isControl()) - exeBranches[thread_number]++; + exeBranches[tid]++; // // Memory operations // if (inst->isMemRef()) { - exeRefs[thread_number]++; + exeRefs[tid]++; if (inst->isLoad()) - exeLoads[thread_number]++; + exeLoads[tid]++; } } @@ -1533,7 +1533,7 @@ template void LWBackEnd::updateComInstStats(DynInstPtr &inst) { - unsigned tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; // keep an instruction count thread->numInst++; diff --git a/src/cpu/ozone/null_predictor.hh b/src/cpu/ozone/null_predictor.hh index 68bb7cd52..44d2f5475 100644 --- a/src/cpu/ozone/null_predictor.hh +++ b/src/cpu/ozone/null_predictor.hh @@ -68,7 +68,7 @@ class NullPredictor * @param tid The thread id. * @return Returns if the branch is taken or not. */ - bool predict(DynInstPtr &inst, Addr &PC, unsigned tid) + bool predict(DynInstPtr &inst, Addr &PC, ThreadID tid) { return false; } /** @@ -77,7 +77,7 @@ class NullPredictor * @param done_sn The sequence number to commit any older updates up until. * @param tid The thread id. */ - void update(const InstSeqNum &done_sn, unsigned tid) { } + void update(const InstSeqNum &done_sn, ThreadID tid) { } /** * Squashes all outstanding updates until a given sequence number. @@ -85,7 +85,7 @@ class NullPredictor * until. * @param tid The thread id. */ - void squash(const InstSeqNum &squashed_sn, unsigned tid) { } + void squash(const InstSeqNum &squashed_sn, ThreadID tid) { } /** * Squashes all outstanding updates until a given sequence number, and @@ -97,7 +97,7 @@ class NullPredictor * @param tid The thread id. */ void squash(const InstSeqNum &squashed_sn, const Addr &corr_target, - bool actually_taken, unsigned tid) + bool actually_taken, ThreadID tid) { } }; diff --git a/src/cpu/ozone/simple_cpu_builder.cc b/src/cpu/ozone/simple_cpu_builder.cc index fa83e2531..2fc7c10d0 100644 --- a/src/cpu/ozone/simple_cpu_builder.cc +++ b/src/cpu/ozone/simple_cpu_builder.cc @@ -62,11 +62,11 @@ SimpleOzoneCPUParams::create() #if FULL_SYSTEM // Full-system only supports a single thread for the moment. - int actual_num_threads = 1; + ThreadID actual_num_threads = 1; #else // In non-full-system mode, we infer the number of threads from // the workload if it's not explicitly specified. - int actual_num_threads = + ThreadID actual_num_threads = numThreads.isValid() ? numThreads : workload.size(); if (workload.size() == 0) { diff --git a/src/cpu/simple/atomic.cc b/src/cpu/simple/atomic.cc index 045b80c80..2ec56c0f2 100644 --- a/src/cpu/simple/atomic.cc +++ b/src/cpu/simple/atomic.cc @@ -80,7 +80,8 @@ AtomicSimpleCPU::init() { BaseCPU::init(); #if FULL_SYSTEM - for (int i = 0; i < threadContexts.size(); ++i) { + ThreadID size = threadContexts.size(); + for (ThreadID i = 0; i < size; ++i) { ThreadContext *tc = threadContexts[i]; // initialize CPU, including PC @@ -227,7 +228,8 @@ AtomicSimpleCPU::takeOverFrom(BaseCPU *oldCPU) // if any of this CPU's ThreadContexts are active, mark the CPU as // running and schedule its tick event. - for (int i = 0; i < threadContexts.size(); ++i) { + ThreadID size = threadContexts.size(); + for (ThreadID i = 0; i < size; ++i) { ThreadContext *tc = threadContexts[i]; if (tc->status() == ThreadContext::Active && _status != Running) { _status = Running; diff --git a/src/cpu/simple/base.cc b/src/cpu/simple/base.cc index 5988f0e7e..61d034f31 100644 --- a/src/cpu/simple/base.cc +++ b/src/cpu/simple/base.cc @@ -200,7 +200,7 @@ BaseSimpleCPU::unserialize(Checkpoint *cp, const string §ion) } void -change_thread_state(int thread_number, int activate, int priority) +change_thread_state(ThreadID tid, int activate, int priority) { } diff --git a/src/cpu/simple/base.hh b/src/cpu/simple/base.hh index e80606388..4e71d677e 100644 --- a/src/cpu/simple/base.hh +++ b/src/cpu/simple/base.hh @@ -399,13 +399,14 @@ class BaseSimpleCPU : public BaseCPU thread->setStCondFailures(sc_failures); } - MiscReg readRegOtherThread(int regIdx, int tid = -1) + MiscReg readRegOtherThread(int regIdx, ThreadID tid = InvalidThreadID) { panic("Simple CPU models do not support multithreaded " "register access.\n"); } - void setRegOtherThread(int regIdx, const MiscReg &val, int tid = -1) + void setRegOtherThread(int regIdx, const MiscReg &val, + ThreadID tid = InvalidThreadID) { panic("Simple CPU models do not support multithreaded " "register access.\n"); diff --git a/src/cpu/simple/timing.cc b/src/cpu/simple/timing.cc index 590ba6b2d..6666f6f9d 100644 --- a/src/cpu/simple/timing.cc +++ b/src/cpu/simple/timing.cc @@ -437,13 +437,13 @@ TimingSimpleCPU::read(Addr addr, T &data, unsigned flags) { Fault fault; const int asid = 0; - const int thread_id = 0; + const ThreadID tid = 0; const Addr pc = thread->readPC(); int block_size = dcachePort.peerBlockSize(); int data_size = sizeof(T); RequestPtr req = new Request(asid, addr, data_size, - flags, pc, _cpuId, thread_id); + flags, pc, _cpuId, tid); Addr split_addr = roundDown(addr + data_size - 1, block_size); assert(split_addr <= addr || split_addr - addr < block_size); @@ -555,13 +555,13 @@ Fault TimingSimpleCPU::write(T data, Addr addr, unsigned flags, uint64_t *res) { const int asid = 0; - const int thread_id = 0; + const ThreadID tid = 0; const Addr pc = thread->readPC(); int block_size = dcachePort.peerBlockSize(); int data_size = sizeof(T); RequestPtr req = new Request(asid, addr, data_size, - flags, pc, _cpuId, thread_id); + flags, pc, _cpuId, tid); Addr split_addr = roundDown(addr + data_size - 1, block_size); assert(split_addr <= addr || split_addr - addr < block_size); diff --git a/src/cpu/simple_thread.hh b/src/cpu/simple_thread.hh index 35f74a67e..08dd45640 100644 --- a/src/cpu/simple_thread.hh +++ b/src/cpu/simple_thread.hh @@ -337,22 +337,26 @@ class SimpleThread : public ThreadState regs.setNextNPC(val); } - MiscReg readMiscRegNoEffect(int misc_reg, unsigned tid = 0) + MiscReg + readMiscRegNoEffect(int misc_reg, ThreadID tid = 0) { return regs.readMiscRegNoEffect(misc_reg); } - MiscReg readMiscReg(int misc_reg, unsigned tid = 0) + MiscReg + readMiscReg(int misc_reg, ThreadID tid = 0) { return regs.readMiscReg(misc_reg, tc); } - void setMiscRegNoEffect(int misc_reg, const MiscReg &val, unsigned tid = 0) + void + setMiscRegNoEffect(int misc_reg, const MiscReg &val, ThreadID tid = 0) { return regs.setMiscRegNoEffect(misc_reg, val); } - void setMiscReg(int misc_reg, const MiscReg &val, unsigned tid = 0) + void + setMiscReg(int misc_reg, const MiscReg &val, ThreadID tid = 0) { return regs.setMiscReg(misc_reg, val, tc); } diff --git a/src/cpu/smt.hh b/src/cpu/smt.hh index cac1a8fd5..d90a1833a 100644 --- a/src/cpu/smt.hh +++ b/src/cpu/smt.hh @@ -52,10 +52,10 @@ extern int maxThreadsPerCPU; /** * Changes the status and priority of the thread with the given number. - * @param thread_number The thread to change. + * @param tid The thread to change. * @param activate The new active status. * @param priority The new priority. */ -void change_thread_state(int thread_number, int activate, int priority); +void change_thread_state(ThreadID tid, int activate, int priority); #endif // __SMT_HH__ diff --git a/src/cpu/thread_context.hh b/src/cpu/thread_context.hh index 51ca3cca6..3e37572d8 100644 --- a/src/cpu/thread_context.hh +++ b/src/cpu/thread_context.hh @@ -234,9 +234,16 @@ class ThreadContext virtual void setMiscReg(int misc_reg, const MiscReg &val) = 0; - virtual uint64_t readRegOtherThread(int misc_reg, unsigned tid) { return 0; } + virtual uint64_t + readRegOtherThread(int misc_reg, ThreadID tid) + { + return 0; + } - virtual void setRegOtherThread(int misc_reg, const MiscReg &val, unsigned tid) { }; + virtual void + setRegOtherThread(int misc_reg, const MiscReg &val, ThreadID tid) + { + } // Also not necessarily the best location for these two. Hopefully will go // away once we decide upon where st cond failures goes. diff --git a/src/cpu/thread_state.cc b/src/cpu/thread_state.cc index 09af2725a..d9d83fb00 100644 --- a/src/cpu/thread_state.cc +++ b/src/cpu/thread_state.cc @@ -43,9 +43,9 @@ #endif #if FULL_SYSTEM -ThreadState::ThreadState(BaseCPU *cpu, int _tid) +ThreadState::ThreadState(BaseCPU *cpu, ThreadID _tid) #else -ThreadState::ThreadState(BaseCPU *cpu, int _tid, +ThreadState::ThreadState(BaseCPU *cpu, ThreadID _tid, Process *_process, short _asid) #endif : numInst(0), numLoad(0), _status(ThreadContext::Halted), diff --git a/src/cpu/thread_state.hh b/src/cpu/thread_state.hh index dbae85998..525ecca86 100644 --- a/src/cpu/thread_state.hh +++ b/src/cpu/thread_state.hh @@ -66,9 +66,9 @@ struct ThreadState { typedef ThreadContext::Status Status; #if FULL_SYSTEM - ThreadState(BaseCPU *cpu, int _tid); + ThreadState(BaseCPU *cpu, ThreadID _tid); #else - ThreadState(BaseCPU *cpu, int _tid, Process *_process, short _asid); + ThreadState(BaseCPU *cpu, ThreadID _tid, Process *_process, short _asid); #endif ~ThreadState(); @@ -83,9 +83,9 @@ struct ThreadState { void setContextId(int id) { _contextId = id; } - void setThreadId(int id) { _threadId = id; } + void setThreadId(ThreadID id) { _threadId = id; } - int threadId() { return _threadId; } + ThreadID threadId() { return _threadId; } Tick readLastActivate() { return lastActivate; } @@ -176,7 +176,7 @@ struct ThreadState { int _contextId; // Index of hardware thread context on the CPU that this represents. - int _threadId; + ThreadID _threadId; public: /** Last time activate was called on this thread. */ diff --git a/src/mem/cache/mshr.cc b/src/mem/cache/mshr.cc index 726253c62..100e6126a 100644 --- a/src/mem/cache/mshr.cc +++ b/src/mem/cache/mshr.cc @@ -51,7 +51,7 @@ MSHR::MSHR() { inService = false; ntargets = 0; - threadNum = -1; + threadNum = InvalidThreadID; targets = new TargetList(); deferredTargets = new TargetList(); } diff --git a/src/mem/cache/mshr.hh b/src/mem/cache/mshr.hh index bed7012b0..13395d314 100644 --- a/src/mem/cache/mshr.hh +++ b/src/mem/cache/mshr.hh @@ -138,7 +138,7 @@ class MSHR : public Packet::SenderState, public Printable bool pendingShared; /** Thread number of the miss. */ - short threadNum; + ThreadID threadNum; /** The number of currently allocated targets. */ short ntargets; diff --git a/src/mem/request.hh b/src/mem/request.hh index 3915fa683..901a24f47 100644 --- a/src/mem/request.hh +++ b/src/mem/request.hh @@ -189,7 +189,7 @@ class Request : public FastAlloc } Request(int asid, Addr vaddr, int size, Flags flags, Addr pc, - int cid, int tid) + int cid, ThreadID tid) { setVirt(asid, vaddr, size, flags, pc); setThreadContext(cid, tid); @@ -201,10 +201,10 @@ class Request : public FastAlloc * Set up CPU and thread numbers. */ void - setThreadContext(int context_id, int thread_id) + setThreadContext(int context_id, ThreadID tid) { _contextId = context_id; - _threadId = thread_id; + _threadId = tid; privateFlags.set(VALID_CONTEXT_ID|VALID_THREAD_ID); } diff --git a/src/mem/ruby/system/CacheMemory.hh b/src/mem/ruby/system/CacheMemory.hh index a8306c06f..6908b9168 100644 --- a/src/mem/ruby/system/CacheMemory.hh +++ b/src/mem/ruby/system/CacheMemory.hh @@ -200,9 +200,9 @@ CacheMemory::CacheMemory(AbstractChip* chip_ptr, int numSetBits, // cout << "Before setting trans address list size" << endl; //create a trans address for each SMT thread // m_trans_address_list.setSize(numThreads); -// for(int i=0; i < numThreads; ++i){ -// cout << "Setting list size for list " << i << endl; -// m_trans_address_list[i].setSize(30); +// for(ThreadID tid = 0; tid < numThreads; ++tid){ +// cout << "Setting list size for list " << tid << endl; +// m_trans_address_list[tid].setSize(30); // } //cout << "CacheMemory constructor finished" << endl; } diff --git a/src/sim/system.hh b/src/sim/system.hh index e1c30490b..ab88160e2 100644 --- a/src/sim/system.hh +++ b/src/sim/system.hh @@ -89,7 +89,7 @@ class System : public SimObject std::vector threadContexts; int _numContexts; - ThreadContext *getThreadContext(int tid) + ThreadContext *getThreadContext(ThreadID tid) { return threadContexts[tid]; }