2005-05-27 05:30:12 +02:00
|
|
|
/*
|
2010-08-23 18:18:40 +02:00
|
|
|
* Copyright (c) 2010 ARM Limited
|
|
|
|
* All rights reserved.
|
|
|
|
*
|
|
|
|
* The license below extends only to copyright in the software and shall
|
|
|
|
* not be construed as granting a license to any other intellectual
|
|
|
|
* property including but not limited to intellectual property relating
|
|
|
|
* to a hardware implementation of the functionality of the software
|
|
|
|
* licensed hereunder. You may use the software subject to the license
|
|
|
|
* terms below provided that you ensure that this notice is replicated
|
|
|
|
* unmodified and in its entirety in all distributions of the software,
|
|
|
|
* modified or unmodified, in source code or in binary form.
|
|
|
|
*
|
2006-05-19 21:44:03 +02:00
|
|
|
* Copyright (c) 2004-2006 The Regents of The University of Michigan
|
2005-05-27 05:30:12 +02:00
|
|
|
* All rights reserved.
|
|
|
|
*
|
|
|
|
* Redistribution and use in source and binary forms, with or without
|
|
|
|
* modification, are permitted provided that the following conditions are
|
|
|
|
* met: redistributions of source code must retain the above copyright
|
|
|
|
* notice, this list of conditions and the following disclaimer;
|
|
|
|
* redistributions in binary form must reproduce the above copyright
|
|
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
|
|
* documentation and/or other materials provided with the distribution;
|
|
|
|
* neither the name of the copyright holders nor the names of its
|
|
|
|
* contributors may be used to endorse or promote products derived from
|
|
|
|
* this software without specific prior written permission.
|
|
|
|
*
|
|
|
|
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
|
|
|
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
|
|
|
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
|
|
|
|
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
|
|
|
|
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
|
|
|
|
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
|
|
|
|
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
|
|
|
|
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
|
|
|
|
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
|
|
|
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
|
|
|
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
2006-06-01 01:26:56 +02:00
|
|
|
*
|
|
|
|
* Authors: Kevin Lim
|
2005-05-27 05:30:12 +02:00
|
|
|
*/
|
|
|
|
|
2004-08-20 20:54:07 +02:00
|
|
|
// @todo: Fix the instantaneous communication among all the stages within
|
|
|
|
// iew. There's a clear delay between issue and execute, yet backwards
|
2005-05-27 05:30:12 +02:00
|
|
|
// communication happens simultaneously.
|
2004-08-20 20:54:07 +02:00
|
|
|
|
|
|
|
#include <queue>
|
|
|
|
|
2011-04-15 19:44:14 +02:00
|
|
|
#include "arch/utility.hh"
|
2009-09-23 17:34:21 +02:00
|
|
|
#include "config/the_isa.hh"
|
2006-04-23 00:26:48 +02:00
|
|
|
#include "cpu/o3/fu_pool.hh"
|
2005-06-05 02:50:10 +02:00
|
|
|
#include "cpu/o3/iew.hh"
|
2011-04-15 19:44:06 +02:00
|
|
|
#include "cpu/timebuf.hh"
|
2011-04-15 19:44:32 +02:00
|
|
|
#include "debug/Activity.hh"
|
|
|
|
#include "debug/Decode.hh"
|
|
|
|
#include "debug/IEW.hh"
|
2008-08-11 21:22:16 +02:00
|
|
|
#include "params/DerivO3CPU.hh"
|
2004-08-20 20:54:07 +02:00
|
|
|
|
2009-05-26 18:23:13 +02:00
|
|
|
using namespace std;
|
|
|
|
|
2005-05-03 16:56:47 +02:00
|
|
|
template<class Impl>
|
2008-08-11 21:22:16 +02:00
|
|
|
DefaultIEW<Impl>::DefaultIEW(O3CPU *_cpu, DerivO3CPUParams *params)
|
2006-07-10 21:40:28 +02:00
|
|
|
: issueToExecQueue(params->backComSize, params->forwardComSize),
|
2007-04-04 21:38:59 +02:00
|
|
|
cpu(_cpu),
|
|
|
|
instQueue(_cpu, this, params),
|
|
|
|
ldstQueue(_cpu, this, params),
|
2006-04-23 00:26:48 +02:00
|
|
|
fuPool(params->fuPool),
|
|
|
|
commitToIEWDelay(params->commitToIEWDelay),
|
|
|
|
renameToIEWDelay(params->renameToIEWDelay),
|
|
|
|
issueToExecuteDelay(params->issueToExecuteDelay),
|
2006-07-05 21:51:36 +02:00
|
|
|
dispatchWidth(params->dispatchWidth),
|
2006-04-23 00:26:48 +02:00
|
|
|
issueWidth(params->issueWidth),
|
2006-07-05 21:51:36 +02:00
|
|
|
wbOutstanding(0),
|
|
|
|
wbWidth(params->wbWidth),
|
2008-08-11 21:22:16 +02:00
|
|
|
numThreads(params->numThreads),
|
2006-05-04 17:36:20 +02:00
|
|
|
switchedOut(false)
|
2006-04-23 00:26:48 +02:00
|
|
|
{
|
|
|
|
_status = Active;
|
|
|
|
exeStatus = Running;
|
|
|
|
wbStatus = Idle;
|
2004-08-20 20:54:07 +02:00
|
|
|
|
|
|
|
// Setup wire to read instructions coming from issue.
|
|
|
|
fromIssue = issueToExecQueue.getWire(-issueToExecuteDelay);
|
|
|
|
|
|
|
|
// Instruction queue needs the queue between issue and execute.
|
|
|
|
instQueue.setIssueToExecuteQueue(&issueToExecQueue);
|
2005-05-03 16:56:47 +02:00
|
|
|
|
2009-05-26 18:23:13 +02:00
|
|
|
for (ThreadID tid = 0; tid < numThreads; tid++) {
|
|
|
|
dispatchStatus[tid] = Running;
|
|
|
|
stalls[tid].commit = false;
|
|
|
|
fetchRedirect[tid] = false;
|
2006-04-23 00:26:48 +02:00
|
|
|
}
|
|
|
|
|
2006-07-05 21:51:36 +02:00
|
|
|
wbMax = wbWidth * params->wbDepth;
|
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
updateLSQNextCycle = false;
|
|
|
|
|
2006-07-05 21:51:36 +02:00
|
|
|
ableToIssue = true;
|
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
skidBufferMax = (3 * (renameToIEWDelay * params->renameWidth)) + issueWidth;
|
|
|
|
}
|
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
std::string
|
|
|
|
DefaultIEW<Impl>::name() const
|
|
|
|
{
|
|
|
|
return cpu->name() + ".iew";
|
2004-08-20 20:54:07 +02:00
|
|
|
}
|
|
|
|
|
2005-05-03 16:56:47 +02:00
|
|
|
template <class Impl>
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
void
|
2006-04-23 00:26:48 +02:00
|
|
|
DefaultIEW<Impl>::regStats()
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
{
|
2006-04-24 23:06:00 +02:00
|
|
|
using namespace Stats;
|
|
|
|
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
instQueue.regStats();
|
2006-06-14 04:35:05 +02:00
|
|
|
ldstQueue.regStats();
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
|
|
|
iewIdleCycles
|
|
|
|
.name(name() + ".iewIdleCycles")
|
|
|
|
.desc("Number of cycles IEW is idle");
|
|
|
|
|
|
|
|
iewSquashCycles
|
|
|
|
.name(name() + ".iewSquashCycles")
|
|
|
|
.desc("Number of cycles IEW is squashing");
|
|
|
|
|
|
|
|
iewBlockCycles
|
|
|
|
.name(name() + ".iewBlockCycles")
|
|
|
|
.desc("Number of cycles IEW is blocking");
|
|
|
|
|
|
|
|
iewUnblockCycles
|
|
|
|
.name(name() + ".iewUnblockCycles")
|
|
|
|
.desc("Number of cycles IEW is unblocking");
|
|
|
|
|
|
|
|
iewDispatchedInsts
|
|
|
|
.name(name() + ".iewDispatchedInsts")
|
|
|
|
.desc("Number of instructions dispatched to IQ");
|
|
|
|
|
|
|
|
iewDispSquashedInsts
|
|
|
|
.name(name() + ".iewDispSquashedInsts")
|
|
|
|
.desc("Number of squashed instructions skipped by dispatch");
|
|
|
|
|
|
|
|
iewDispLoadInsts
|
|
|
|
.name(name() + ".iewDispLoadInsts")
|
|
|
|
.desc("Number of dispatched load instructions");
|
|
|
|
|
|
|
|
iewDispStoreInsts
|
|
|
|
.name(name() + ".iewDispStoreInsts")
|
|
|
|
.desc("Number of dispatched store instructions");
|
|
|
|
|
|
|
|
iewDispNonSpecInsts
|
|
|
|
.name(name() + ".iewDispNonSpecInsts")
|
|
|
|
.desc("Number of dispatched non-speculative instructions");
|
|
|
|
|
|
|
|
iewIQFullEvents
|
|
|
|
.name(name() + ".iewIQFullEvents")
|
|
|
|
.desc("Number of times the IQ has become full, causing a stall");
|
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
iewLSQFullEvents
|
|
|
|
.name(name() + ".iewLSQFullEvents")
|
|
|
|
.desc("Number of times the LSQ has become full, causing a stall");
|
|
|
|
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
memOrderViolationEvents
|
|
|
|
.name(name() + ".memOrderViolationEvents")
|
|
|
|
.desc("Number of memory order violations");
|
|
|
|
|
|
|
|
predictedTakenIncorrect
|
|
|
|
.name(name() + ".predictedTakenIncorrect")
|
|
|
|
.desc("Number of branches that were predicted taken incorrectly");
|
2006-04-23 00:26:48 +02:00
|
|
|
|
|
|
|
predictedNotTakenIncorrect
|
|
|
|
.name(name() + ".predictedNotTakenIncorrect")
|
|
|
|
.desc("Number of branches that were predicted not taken incorrectly");
|
|
|
|
|
|
|
|
branchMispredicts
|
|
|
|
.name(name() + ".branchMispredicts")
|
|
|
|
.desc("Number of branch mispredicts detected at execute");
|
|
|
|
|
|
|
|
branchMispredicts = predictedTakenIncorrect + predictedNotTakenIncorrect;
|
2006-04-24 23:06:00 +02:00
|
|
|
|
2006-06-14 04:35:05 +02:00
|
|
|
iewExecutedInsts
|
2006-08-02 18:05:34 +02:00
|
|
|
.name(name() + ".iewExecutedInsts")
|
2006-06-14 04:35:05 +02:00
|
|
|
.desc("Number of executed instructions");
|
|
|
|
|
|
|
|
iewExecLoadInsts
|
2009-05-26 18:23:13 +02:00
|
|
|
.init(cpu->numThreads)
|
2006-08-02 18:05:34 +02:00
|
|
|
.name(name() + ".iewExecLoadInsts")
|
2006-06-14 04:35:05 +02:00
|
|
|
.desc("Number of load instructions executed")
|
|
|
|
.flags(total);
|
|
|
|
|
|
|
|
iewExecSquashedInsts
|
2006-08-02 18:05:34 +02:00
|
|
|
.name(name() + ".iewExecSquashedInsts")
|
2006-06-14 04:35:05 +02:00
|
|
|
.desc("Number of squashed instructions skipped in execute");
|
|
|
|
|
|
|
|
iewExecutedSwp
|
2009-05-26 18:23:13 +02:00
|
|
|
.init(cpu->numThreads)
|
2011-04-20 03:45:21 +02:00
|
|
|
.name(name() + ".exec_swp")
|
2006-04-24 23:06:00 +02:00
|
|
|
.desc("number of swp insts executed")
|
2006-06-14 04:35:05 +02:00
|
|
|
.flags(total);
|
2006-04-24 23:06:00 +02:00
|
|
|
|
2006-06-14 04:35:05 +02:00
|
|
|
iewExecutedNop
|
2009-05-26 18:23:13 +02:00
|
|
|
.init(cpu->numThreads)
|
2011-04-20 03:45:21 +02:00
|
|
|
.name(name() + ".exec_nop")
|
2006-04-24 23:06:00 +02:00
|
|
|
.desc("number of nop insts executed")
|
2006-06-14 04:35:05 +02:00
|
|
|
.flags(total);
|
2006-04-24 23:06:00 +02:00
|
|
|
|
2006-06-14 04:35:05 +02:00
|
|
|
iewExecutedRefs
|
2009-05-26 18:23:13 +02:00
|
|
|
.init(cpu->numThreads)
|
2011-04-20 03:45:21 +02:00
|
|
|
.name(name() + ".exec_refs")
|
2006-04-24 23:06:00 +02:00
|
|
|
.desc("number of memory reference insts executed")
|
2006-06-14 04:35:05 +02:00
|
|
|
.flags(total);
|
2006-04-24 23:06:00 +02:00
|
|
|
|
2006-06-14 04:35:05 +02:00
|
|
|
iewExecutedBranches
|
2009-05-26 18:23:13 +02:00
|
|
|
.init(cpu->numThreads)
|
2011-04-20 03:45:21 +02:00
|
|
|
.name(name() + ".exec_branches")
|
2006-04-24 23:06:00 +02:00
|
|
|
.desc("Number of branches executed")
|
2006-06-14 04:35:05 +02:00
|
|
|
.flags(total);
|
2006-04-24 23:06:00 +02:00
|
|
|
|
|
|
|
iewExecStoreInsts
|
2011-04-20 03:45:21 +02:00
|
|
|
.name(name() + ".exec_stores")
|
2006-04-24 23:06:00 +02:00
|
|
|
.desc("Number of stores executed")
|
2006-06-14 04:35:05 +02:00
|
|
|
.flags(total);
|
|
|
|
iewExecStoreInsts = iewExecutedRefs - iewExecLoadInsts;
|
|
|
|
|
|
|
|
iewExecRate
|
2011-04-20 03:45:21 +02:00
|
|
|
.name(name() + ".exec_rate")
|
2006-06-14 04:35:05 +02:00
|
|
|
.desc("Inst execution rate")
|
|
|
|
.flags(total);
|
|
|
|
|
|
|
|
iewExecRate = iewExecutedInsts / cpu->numCycles;
|
2006-04-24 23:06:00 +02:00
|
|
|
|
|
|
|
iewInstsToCommit
|
2009-05-26 18:23:13 +02:00
|
|
|
.init(cpu->numThreads)
|
2011-04-20 03:45:21 +02:00
|
|
|
.name(name() + ".wb_sent")
|
2006-04-24 23:06:00 +02:00
|
|
|
.desc("cumulative count of insts sent to commit")
|
2006-06-14 04:35:05 +02:00
|
|
|
.flags(total);
|
2006-04-24 23:06:00 +02:00
|
|
|
|
2006-05-19 21:44:03 +02:00
|
|
|
writebackCount
|
2009-05-26 18:23:13 +02:00
|
|
|
.init(cpu->numThreads)
|
2011-04-20 03:45:21 +02:00
|
|
|
.name(name() + ".wb_count")
|
2006-04-24 23:06:00 +02:00
|
|
|
.desc("cumulative count of insts written-back")
|
2006-06-14 04:35:05 +02:00
|
|
|
.flags(total);
|
2006-04-24 23:06:00 +02:00
|
|
|
|
2006-05-19 21:44:03 +02:00
|
|
|
producerInst
|
2009-05-26 18:23:13 +02:00
|
|
|
.init(cpu->numThreads)
|
2011-04-20 03:45:21 +02:00
|
|
|
.name(name() + ".wb_producers")
|
2006-04-24 23:06:00 +02:00
|
|
|
.desc("num instructions producing a value")
|
2006-06-14 04:35:05 +02:00
|
|
|
.flags(total);
|
2006-04-24 23:06:00 +02:00
|
|
|
|
2006-05-19 21:44:03 +02:00
|
|
|
consumerInst
|
2009-05-26 18:23:13 +02:00
|
|
|
.init(cpu->numThreads)
|
2011-04-20 03:45:21 +02:00
|
|
|
.name(name() + ".wb_consumers")
|
2006-04-24 23:06:00 +02:00
|
|
|
.desc("num instructions consuming a value")
|
2006-06-14 04:35:05 +02:00
|
|
|
.flags(total);
|
2006-04-24 23:06:00 +02:00
|
|
|
|
2006-05-19 21:44:03 +02:00
|
|
|
wbPenalized
|
2009-05-26 18:23:13 +02:00
|
|
|
.init(cpu->numThreads)
|
2011-04-20 03:45:21 +02:00
|
|
|
.name(name() + ".wb_penalized")
|
2006-04-24 23:06:00 +02:00
|
|
|
.desc("number of instrctions required to write to 'other' IQ")
|
2006-06-14 04:35:05 +02:00
|
|
|
.flags(total);
|
2006-04-24 23:06:00 +02:00
|
|
|
|
2006-05-19 21:44:03 +02:00
|
|
|
wbPenalizedRate
|
2011-04-20 03:45:21 +02:00
|
|
|
.name(name() + ".wb_penalized_rate")
|
2006-04-24 23:06:00 +02:00
|
|
|
.desc ("fraction of instructions written-back that wrote to 'other' IQ")
|
2006-06-14 04:35:05 +02:00
|
|
|
.flags(total);
|
2006-04-24 23:06:00 +02:00
|
|
|
|
2006-05-19 21:44:03 +02:00
|
|
|
wbPenalizedRate = wbPenalized / writebackCount;
|
2006-04-24 23:06:00 +02:00
|
|
|
|
2006-05-19 21:44:03 +02:00
|
|
|
wbFanout
|
2011-04-20 03:45:21 +02:00
|
|
|
.name(name() + ".wb_fanout")
|
2006-04-24 23:06:00 +02:00
|
|
|
.desc("average fanout of values written-back")
|
2006-06-14 04:35:05 +02:00
|
|
|
.flags(total);
|
2006-04-24 23:06:00 +02:00
|
|
|
|
2006-05-19 21:44:03 +02:00
|
|
|
wbFanout = producerInst / consumerInst;
|
2006-04-24 23:06:00 +02:00
|
|
|
|
2006-05-19 21:44:03 +02:00
|
|
|
wbRate
|
2011-04-20 03:45:21 +02:00
|
|
|
.name(name() + ".wb_rate")
|
2006-04-24 23:06:00 +02:00
|
|
|
.desc("insts written-back per cycle")
|
2006-06-14 04:35:05 +02:00
|
|
|
.flags(total);
|
2006-05-19 21:44:03 +02:00
|
|
|
wbRate = writebackCount / cpu->numCycles;
|
2006-04-23 00:26:48 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
template<class Impl>
|
|
|
|
void
|
|
|
|
DefaultIEW<Impl>::initStage()
|
|
|
|
{
|
2009-05-26 18:23:13 +02:00
|
|
|
for (ThreadID tid = 0; tid < numThreads; tid++) {
|
2006-04-23 00:26:48 +02:00
|
|
|
toRename->iewInfo[tid].usedIQ = true;
|
|
|
|
toRename->iewInfo[tid].freeIQEntries =
|
|
|
|
instQueue.numFreeEntries(tid);
|
|
|
|
|
|
|
|
toRename->iewInfo[tid].usedLSQ = true;
|
|
|
|
toRename->iewInfo[tid].freeLSQEntries =
|
|
|
|
ldstQueue.numFreeEntries(tid);
|
|
|
|
}
|
|
|
|
|
2006-06-16 23:08:47 +02:00
|
|
|
cpu->activateStage(O3CPU::IEWIdx);
|
2004-08-20 20:54:07 +02:00
|
|
|
}
|
|
|
|
|
2005-05-03 16:56:47 +02:00
|
|
|
template<class Impl>
|
2004-08-20 20:54:07 +02:00
|
|
|
void
|
2006-04-23 00:26:48 +02:00
|
|
|
DefaultIEW<Impl>::setTimeBuffer(TimeBuffer<TimeStruct> *tb_ptr)
|
2004-08-20 20:54:07 +02:00
|
|
|
{
|
|
|
|
timeBuffer = tb_ptr;
|
|
|
|
|
|
|
|
// Setup wire to read information from time buffer, from commit.
|
|
|
|
fromCommit = timeBuffer->getWire(-commitToIEWDelay);
|
|
|
|
|
|
|
|
// Setup wire to write information back to previous stages.
|
|
|
|
toRename = timeBuffer->getWire(0);
|
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
toFetch = timeBuffer->getWire(0);
|
|
|
|
|
2004-08-20 20:54:07 +02:00
|
|
|
// Instruction queue also needs main time buffer.
|
|
|
|
instQueue.setTimeBuffer(tb_ptr);
|
|
|
|
}
|
|
|
|
|
2005-05-03 16:56:47 +02:00
|
|
|
template<class Impl>
|
2004-08-20 20:54:07 +02:00
|
|
|
void
|
2006-04-23 00:26:48 +02:00
|
|
|
DefaultIEW<Impl>::setRenameQueue(TimeBuffer<RenameStruct> *rq_ptr)
|
2004-08-20 20:54:07 +02:00
|
|
|
{
|
|
|
|
renameQueue = rq_ptr;
|
|
|
|
|
|
|
|
// Setup wire to read information from rename queue.
|
|
|
|
fromRename = renameQueue->getWire(-renameToIEWDelay);
|
|
|
|
}
|
|
|
|
|
2005-05-03 16:56:47 +02:00
|
|
|
template<class Impl>
|
2004-08-20 20:54:07 +02:00
|
|
|
void
|
2006-04-23 00:26:48 +02:00
|
|
|
DefaultIEW<Impl>::setIEWQueue(TimeBuffer<IEWStruct> *iq_ptr)
|
2004-08-20 20:54:07 +02:00
|
|
|
{
|
|
|
|
iewQueue = iq_ptr;
|
|
|
|
|
|
|
|
// Setup wire to write instructions to commit.
|
|
|
|
toCommit = iewQueue->getWire(0);
|
|
|
|
}
|
|
|
|
|
2005-05-03 16:56:47 +02:00
|
|
|
template<class Impl>
|
2004-08-20 20:54:07 +02:00
|
|
|
void
|
2009-05-26 18:23:13 +02:00
|
|
|
DefaultIEW<Impl>::setActiveThreads(list<ThreadID> *at_ptr)
|
2006-04-23 00:26:48 +02:00
|
|
|
{
|
|
|
|
activeThreads = at_ptr;
|
|
|
|
|
|
|
|
ldstQueue.setActiveThreads(at_ptr);
|
|
|
|
instQueue.setActiveThreads(at_ptr);
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class Impl>
|
|
|
|
void
|
|
|
|
DefaultIEW<Impl>::setScoreboard(Scoreboard *sb_ptr)
|
2004-08-20 20:54:07 +02:00
|
|
|
{
|
2006-04-23 00:26:48 +02:00
|
|
|
scoreboard = sb_ptr;
|
2004-08-20 20:54:07 +02:00
|
|
|
}
|
|
|
|
|
2006-05-04 17:36:20 +02:00
|
|
|
template <class Impl>
|
2006-07-06 23:57:20 +02:00
|
|
|
bool
|
2006-07-06 19:59:02 +02:00
|
|
|
DefaultIEW<Impl>::drain()
|
2006-05-16 20:06:35 +02:00
|
|
|
{
|
2006-07-06 19:59:02 +02:00
|
|
|
// IEW is ready to drain at any time.
|
|
|
|
cpu->signalDrained();
|
2006-07-06 23:57:20 +02:00
|
|
|
return true;
|
2006-05-16 20:06:35 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
void
|
2006-07-06 19:59:02 +02:00
|
|
|
DefaultIEW<Impl>::resume()
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
void
|
|
|
|
DefaultIEW<Impl>::switchOut()
|
2006-05-04 17:36:20 +02:00
|
|
|
{
|
2006-05-31 17:45:02 +02:00
|
|
|
// Clear any state.
|
2006-05-04 17:36:20 +02:00
|
|
|
switchedOut = true;
|
2006-09-28 06:09:27 +02:00
|
|
|
assert(insts[0].empty());
|
|
|
|
assert(skidBuffer[0].empty());
|
2006-05-16 20:06:35 +02:00
|
|
|
|
2006-05-04 17:36:20 +02:00
|
|
|
instQueue.switchOut();
|
|
|
|
ldstQueue.switchOut();
|
|
|
|
fuPool->switchOut();
|
|
|
|
|
2009-05-26 18:23:13 +02:00
|
|
|
for (ThreadID tid = 0; tid < numThreads; tid++) {
|
|
|
|
while (!insts[tid].empty())
|
|
|
|
insts[tid].pop();
|
|
|
|
while (!skidBuffer[tid].empty())
|
|
|
|
skidBuffer[tid].pop();
|
2006-05-04 17:36:20 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
void
|
|
|
|
DefaultIEW<Impl>::takeOverFrom()
|
|
|
|
{
|
2006-05-31 17:45:02 +02:00
|
|
|
// Reset all state.
|
2006-05-04 17:36:20 +02:00
|
|
|
_status = Active;
|
|
|
|
exeStatus = Running;
|
|
|
|
wbStatus = Idle;
|
|
|
|
switchedOut = false;
|
|
|
|
|
|
|
|
instQueue.takeOverFrom();
|
|
|
|
ldstQueue.takeOverFrom();
|
|
|
|
fuPool->takeOverFrom();
|
|
|
|
|
|
|
|
initStage();
|
|
|
|
cpu->activityThisCycle();
|
|
|
|
|
2009-05-26 18:23:13 +02:00
|
|
|
for (ThreadID tid = 0; tid < numThreads; tid++) {
|
|
|
|
dispatchStatus[tid] = Running;
|
|
|
|
stalls[tid].commit = false;
|
|
|
|
fetchRedirect[tid] = false;
|
2006-05-04 17:36:20 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
updateLSQNextCycle = false;
|
|
|
|
|
2006-07-10 21:40:28 +02:00
|
|
|
for (int i = 0; i < issueToExecQueue.getSize(); ++i) {
|
2006-05-04 17:36:20 +02:00
|
|
|
issueToExecQueue.advance();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
template<class Impl>
|
|
|
|
void
|
2009-05-26 18:23:13 +02:00
|
|
|
DefaultIEW<Impl>::squash(ThreadID tid)
|
2006-04-23 00:26:48 +02:00
|
|
|
{
|
2009-05-26 18:23:13 +02:00
|
|
|
DPRINTF(IEW, "[tid:%i]: Squashing all instructions.\n", tid);
|
2005-05-03 16:56:47 +02:00
|
|
|
|
|
|
|
// Tell the IQ to start squashing.
|
2006-04-23 00:26:48 +02:00
|
|
|
instQueue.squash(tid);
|
2005-05-03 16:56:47 +02:00
|
|
|
|
|
|
|
// Tell the LDSTQ to start squashing.
|
2006-09-01 02:51:30 +02:00
|
|
|
ldstQueue.squash(fromCommit->commitInfo[tid].doneSeqNum, tid);
|
2006-04-23 00:26:48 +02:00
|
|
|
updatedQueues = true;
|
|
|
|
|
|
|
|
// Clear the skid buffer in case it has any data in it.
|
2006-07-23 19:39:42 +02:00
|
|
|
DPRINTF(IEW, "[tid:%i]: Removing skidbuffer instructions until [sn:%i].\n",
|
2007-04-13 15:59:31 +02:00
|
|
|
tid, fromCommit->commitInfo[tid].doneSeqNum);
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2006-07-23 19:39:42 +02:00
|
|
|
while (!skidBuffer[tid].empty()) {
|
2006-04-23 00:26:48 +02:00
|
|
|
if (skidBuffer[tid].front()->isLoad() ||
|
|
|
|
skidBuffer[tid].front()->isStore() ) {
|
|
|
|
toRename->iewInfo[tid].dispatchedToLSQ++;
|
|
|
|
}
|
|
|
|
|
|
|
|
toRename->iewInfo[tid].dispatched++;
|
|
|
|
|
|
|
|
skidBuffer[tid].pop();
|
|
|
|
}
|
|
|
|
|
2006-06-13 01:04:42 +02:00
|
|
|
emptyRenameInsts(tid);
|
2005-05-03 16:56:47 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
template<class Impl>
|
|
|
|
void
|
2009-05-26 18:23:13 +02:00
|
|
|
DefaultIEW<Impl>::squashDueToBranch(DynInstPtr &inst, ThreadID tid)
|
2005-05-03 16:56:47 +02:00
|
|
|
{
|
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
|
|
|
DPRINTF(IEW, "[tid:%i]: Squashing from a specific instruction, PC: %s "
|
|
|
|
"[sn:%i].\n", tid, inst->pcState(), inst->seqNum);
|
2005-05-03 16:56:47 +02:00
|
|
|
|
2011-01-18 23:30:05 +01:00
|
|
|
if (toCommit->squash[tid] == false ||
|
|
|
|
inst->seqNum < toCommit->squashedSeqNum[tid]) {
|
|
|
|
toCommit->squash[tid] = true;
|
|
|
|
toCommit->squashedSeqNum[tid] = inst->seqNum;
|
|
|
|
toCommit->branchTaken[tid] = inst->pcState().branching();
|
2006-07-23 19:39:42 +02:00
|
|
|
|
2011-01-18 23:30:05 +01:00
|
|
|
TheISA::PCState pc = inst->pcState();
|
|
|
|
TheISA::advancePC(pc, inst->staticInst);
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2011-01-18 23:30:05 +01:00
|
|
|
toCommit->pc[tid] = pc;
|
|
|
|
toCommit->mispredictInst[tid] = inst;
|
|
|
|
toCommit->includeSquashInst[tid] = false;
|
|
|
|
|
|
|
|
wroteToTimeBuffer = true;
|
|
|
|
}
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2005-05-03 16:56:47 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
template<class Impl>
|
|
|
|
void
|
2009-05-26 18:23:13 +02:00
|
|
|
DefaultIEW<Impl>::squashDueToMemOrder(DynInstPtr &inst, ThreadID tid)
|
2005-05-03 16:56:47 +02:00
|
|
|
{
|
2011-08-19 22:08:05 +02:00
|
|
|
DPRINTF(IEW, "[tid:%i]: Memory violation, squashing violator and younger "
|
|
|
|
"insts, PC: %s [sn:%i].\n", tid, inst->pcState(), inst->seqNum);
|
|
|
|
// Need to include inst->seqNum in the following comparison to cover the
|
|
|
|
// corner case when a branch misprediction and a memory violation for the
|
|
|
|
// same instruction (e.g. load PC) are detected in the same cycle. In this
|
|
|
|
// case the memory violator should take precedence over the branch
|
|
|
|
// misprediction because it requires the violator itself to be included in
|
|
|
|
// the squash.
|
2011-01-18 23:30:05 +01:00
|
|
|
if (toCommit->squash[tid] == false ||
|
2011-08-19 22:08:05 +02:00
|
|
|
inst->seqNum <= toCommit->squashedSeqNum[tid]) {
|
2011-01-18 23:30:05 +01:00
|
|
|
toCommit->squash[tid] = true;
|
2011-08-19 22:08:05 +02:00
|
|
|
|
2011-01-18 23:30:05 +01:00
|
|
|
toCommit->squashedSeqNum[tid] = inst->seqNum;
|
2011-08-19 22:08:05 +02:00
|
|
|
toCommit->pc[tid] = inst->pcState();
|
2011-03-18 01:20:19 +01:00
|
|
|
toCommit->mispredictInst[tid] = NULL;
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2011-08-19 22:08:05 +02:00
|
|
|
// Must include the memory violator in the squash.
|
|
|
|
toCommit->includeSquashInst[tid] = true;
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2011-01-18 23:30:05 +01:00
|
|
|
wroteToTimeBuffer = true;
|
|
|
|
}
|
2004-08-20 20:54:07 +02:00
|
|
|
}
|
|
|
|
|
2005-05-03 16:56:47 +02:00
|
|
|
template<class Impl>
|
2004-08-20 20:54:07 +02:00
|
|
|
void
|
2009-05-26 18:23:13 +02:00
|
|
|
DefaultIEW<Impl>::squashDueToMemBlocked(DynInstPtr &inst, ThreadID tid)
|
2004-08-20 20:54:07 +02:00
|
|
|
{
|
2006-04-23 00:26:48 +02:00
|
|
|
DPRINTF(IEW, "[tid:%i]: Memory blocked, squashing load and younger insts, "
|
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
|
|
|
"PC: %s [sn:%i].\n", tid, inst->pcState(), inst->seqNum);
|
2011-01-18 23:30:05 +01:00
|
|
|
if (toCommit->squash[tid] == false ||
|
|
|
|
inst->seqNum < toCommit->squashedSeqNum[tid]) {
|
|
|
|
toCommit->squash[tid] = true;
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2011-01-18 23:30:05 +01:00
|
|
|
toCommit->squashedSeqNum[tid] = inst->seqNum;
|
|
|
|
toCommit->pc[tid] = inst->pcState();
|
2011-03-18 01:20:19 +01:00
|
|
|
toCommit->mispredictInst[tid] = NULL;
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2011-01-18 23:30:05 +01:00
|
|
|
// Must include the broadcasted SN in the squash.
|
|
|
|
toCommit->includeSquashInst[tid] = true;
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2011-01-18 23:30:05 +01:00
|
|
|
ldstQueue.setLoadBlockedHandled(tid);
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2011-01-18 23:30:05 +01:00
|
|
|
wroteToTimeBuffer = true;
|
|
|
|
}
|
2006-04-23 00:26:48 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
template<class Impl>
|
|
|
|
void
|
2009-05-26 18:23:13 +02:00
|
|
|
DefaultIEW<Impl>::block(ThreadID tid)
|
2006-04-23 00:26:48 +02:00
|
|
|
{
|
|
|
|
DPRINTF(IEW, "[tid:%u]: Blocking.\n", tid);
|
|
|
|
|
|
|
|
if (dispatchStatus[tid] != Blocked &&
|
|
|
|
dispatchStatus[tid] != Unblocking) {
|
|
|
|
toRename->iewBlock[tid] = true;
|
|
|
|
wroteToTimeBuffer = true;
|
|
|
|
}
|
2004-08-20 20:54:07 +02:00
|
|
|
|
|
|
|
// Add the current inputs to the skid buffer so they can be
|
|
|
|
// reprocessed when this stage unblocks.
|
2006-04-23 00:26:48 +02:00
|
|
|
skidInsert(tid);
|
2004-08-20 20:54:07 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
dispatchStatus[tid] = Blocked;
|
2004-08-20 20:54:07 +02:00
|
|
|
}
|
|
|
|
|
2005-05-03 16:56:47 +02:00
|
|
|
template<class Impl>
|
2006-04-23 00:26:48 +02:00
|
|
|
void
|
2009-05-26 18:23:13 +02:00
|
|
|
DefaultIEW<Impl>::unblock(ThreadID tid)
|
2004-08-20 20:54:07 +02:00
|
|
|
{
|
2006-04-23 00:26:48 +02:00
|
|
|
DPRINTF(IEW, "[tid:%i]: Reading instructions out of the skid "
|
|
|
|
"buffer %u.\n",tid, tid);
|
|
|
|
|
|
|
|
// If the skid bufffer is empty, signal back to previous stages to unblock.
|
|
|
|
// Also switch status to running.
|
|
|
|
if (skidBuffer[tid].empty()) {
|
|
|
|
toRename->iewUnblock[tid] = true;
|
|
|
|
wroteToTimeBuffer = true;
|
|
|
|
DPRINTF(IEW, "[tid:%i]: Done unblocking.\n",tid);
|
|
|
|
dispatchStatus[tid] = Running;
|
2004-08-20 20:54:07 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2005-05-03 16:56:47 +02:00
|
|
|
template<class Impl>
|
2004-08-20 20:54:07 +02:00
|
|
|
void
|
2006-04-23 00:26:48 +02:00
|
|
|
DefaultIEW<Impl>::wakeDependents(DynInstPtr &inst)
|
2004-08-20 20:54:07 +02:00
|
|
|
{
|
2005-05-03 16:56:47 +02:00
|
|
|
instQueue.wakeDependents(inst);
|
2004-08-20 20:54:07 +02:00
|
|
|
}
|
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
template<class Impl>
|
|
|
|
void
|
|
|
|
DefaultIEW<Impl>::rescheduleMemInst(DynInstPtr &inst)
|
|
|
|
{
|
|
|
|
instQueue.rescheduleMemInst(inst);
|
|
|
|
}
|
2004-08-20 20:54:07 +02:00
|
|
|
|
2005-05-03 16:56:47 +02:00
|
|
|
template<class Impl>
|
2004-08-20 20:54:07 +02:00
|
|
|
void
|
2006-04-23 00:26:48 +02:00
|
|
|
DefaultIEW<Impl>::replayMemInst(DynInstPtr &inst)
|
2004-08-20 20:54:07 +02:00
|
|
|
{
|
2006-04-23 00:26:48 +02:00
|
|
|
instQueue.replayMemInst(inst);
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class Impl>
|
|
|
|
void
|
|
|
|
DefaultIEW<Impl>::instToCommit(DynInstPtr &inst)
|
|
|
|
{
|
2006-10-08 06:53:41 +02:00
|
|
|
// This function should not be called after writebackInsts in a
|
|
|
|
// single cycle. That will cause problems with an instruction
|
|
|
|
// being added to the queue to commit without being processed by
|
|
|
|
// writebackInsts prior to being sent to commit.
|
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
// First check the time slot that this instruction will write
|
|
|
|
// to. If there are free write ports at the time, then go ahead
|
|
|
|
// and write the instruction to that time. If there are not,
|
|
|
|
// keep looking back to see where's the first time there's a
|
2006-05-19 21:44:03 +02:00
|
|
|
// free slot.
|
2006-04-23 00:26:48 +02:00
|
|
|
while ((*iewQueue)[wbCycle].insts[wbNumInst]) {
|
|
|
|
++wbNumInst;
|
2006-07-05 21:51:36 +02:00
|
|
|
if (wbNumInst == wbWidth) {
|
2006-04-23 00:26:48 +02:00
|
|
|
++wbCycle;
|
|
|
|
wbNumInst = 0;
|
|
|
|
}
|
|
|
|
|
2006-08-02 18:05:34 +02:00
|
|
|
assert((wbCycle * wbWidth + wbNumInst) <= wbMax);
|
2006-04-23 00:26:48 +02:00
|
|
|
}
|
Update to make multiple instruction issue and different latencies work.
Also change to ref counted DynInst.
SConscript:
Add branch predictor, BTB, load store queue, and storesets.
arch/isa_parser.py:
Specify the template parameter for AlphaDynInst
base/traceflags.py:
Add load store queue, store set, and mem dependence unit to the
list of trace flags.
cpu/base_dyn_inst.cc:
Change formating, add in debug statement.
cpu/base_dyn_inst.hh:
Change DynInst to be RefCounted, add flag to clear whether or not this
instruction can commit. This is likely to be removed in the future.
cpu/beta_cpu/alpha_dyn_inst.cc:
AlphaDynInst has been changed to be templated, so now this CC file
is just used to force instantiations of AlphaDynInst.
cpu/beta_cpu/alpha_dyn_inst.hh:
Changed AlphaDynInst to be templated on Impl. Removed some unnecessary
functions.
cpu/beta_cpu/alpha_full_cpu.cc:
AlphaFullCPU has been changed to be templated, so this CC file is now
just used to force instantation of AlphaFullCPU.
cpu/beta_cpu/alpha_full_cpu.hh:
Change AlphaFullCPU to be templated on Impl.
cpu/beta_cpu/alpha_impl.hh:
Update it to reflect AlphaDynInst and AlphaFullCPU being templated
on Impl. Also removed time buffers from here, as they are really
a part of the CPU and are thus in the CPU policy now.
cpu/beta_cpu/alpha_params.hh:
Make AlphaSimpleParams inherit from the BaseFullCPU so that it doesn't
need to specifically declare any parameters that are already in the
BaseFullCPU.
cpu/beta_cpu/comm.hh:
Changed the structure of the time buffer communication structs. Now
they include the size of the packet of instructions it is sending.
Added some parameters to the backwards communication struct, mainly
for squashing.
cpu/beta_cpu/commit.hh:
Update typenames to reflect change in location of time buffer structs.
Update DynInst to DynInstPtr (it is refcounted now).
cpu/beta_cpu/commit_impl.hh:
Formatting changes mainly. Also sends back proper information
on branch mispredicts so that the bpred unit can update itself.
Updated behavior for non-speculative instructions (stores, any
other non-spec instructions): once they reach the head of the ROB,
the ROB signals back to the IQ that it can go ahead and issue the
non-speculative instruction. The instruction itself is updated so that
commit won't try to commit it again until it is done executing.
cpu/beta_cpu/cpu_policy.hh:
Added branch prediction unit, mem dependence prediction unit, load
store queue. Moved time buffer structs from AlphaSimpleImpl to here.
cpu/beta_cpu/decode.hh:
Changed typedefs to reflect change in location of time buffer structs
and also the change from DynInst to ref counted DynInstPtr.
cpu/beta_cpu/decode_impl.hh:
Continues to buffer instructions even while unblocking now. Changed
how it loops through groups of instructions so it can properly block
during the middle of a group of instructions.
cpu/beta_cpu/fetch.hh:
Changed typedefs to reflect change in location of time buffer structs
and the change to ref counted DynInsts. Also added in branch
brediction unit.
cpu/beta_cpu/fetch_impl.hh:
Add in branch prediction. Changed how fetch checks inputs and its
current state to make for easier logic.
cpu/beta_cpu/free_list.cc:
Changed int regs and float regs to logically use one flat namespace.
Future change will be moving them to a single scoreboard to conserve
space.
cpu/beta_cpu/free_list.hh:
Mostly debugging statements. Might be removed for performance in future.
cpu/beta_cpu/full_cpu.cc:
Added in some debugging statements. Updated BaseFullCPU to take
a params object.
cpu/beta_cpu/full_cpu.hh:
Added params class within BaseCPU that other param classes will be
able to inherit from. Updated typedefs to reflect change in location
of time buffer structs and ref counted DynInst.
cpu/beta_cpu/iew.hh:
Updated typedefs to reflect change in location of time buffer structs
and use of ref counted DynInsts.
cpu/beta_cpu/iew_impl.hh:
Added in load store queue, updated iew to be able to execute non-
speculative instructions, instead of having them execute in commit.
cpu/beta_cpu/inst_queue.hh:
Updated change to ref counted DynInsts. Changed inst queue to hold
non-speculative instructions as well, which are issued only when
commit signals backwards that a nonspeculative instruction is at
the head of the ROB.
cpu/beta_cpu/inst_queue_impl.hh:
Updated to allow for non-speculative instructions to be in the inst
queue. Also added some debug functions.
cpu/beta_cpu/regfile.hh:
Added debugging statements, changed formatting.
cpu/beta_cpu/rename.hh:
Updated typedefs, added some functions to clean up code.
cpu/beta_cpu/rename_impl.hh:
Moved some code into functions to make it easier to read.
cpu/beta_cpu/rename_map.cc:
Changed int and float reg behavior to use a single flat namespace. In
the future, the rename maps can be combined to a single rename map to
save space.
cpu/beta_cpu/rename_map.hh:
Added destructor.
cpu/beta_cpu/rob.hh:
Updated it with change from DynInst to ref counted DynInst.
cpu/beta_cpu/rob_impl.hh:
Formatting, updated to use ref counted DynInst.
cpu/static_inst.hh:
Updated forward declaration for AlphaDynInst now that it is templated.
--HG--
extra : convert_revision : 1045f240ee9b6a4bd368e1806aca029ebbdc6dd3
2004-09-23 20:06:03 +02:00
|
|
|
|
2006-08-02 18:05:34 +02:00
|
|
|
DPRINTF(IEW, "Current wb cycle: %i, width: %i, numInst: %i\nwbActual:%i\n",
|
|
|
|
wbCycle, wbWidth, wbNumInst, wbCycle * wbWidth + wbNumInst);
|
2006-04-23 00:26:48 +02:00
|
|
|
// Add finished instruction to queue to commit.
|
|
|
|
(*iewQueue)[wbCycle].insts[wbNumInst] = inst;
|
|
|
|
(*iewQueue)[wbCycle].size++;
|
2004-08-20 20:54:07 +02:00
|
|
|
}
|
|
|
|
|
2005-05-03 16:56:47 +02:00
|
|
|
template <class Impl>
|
2006-04-23 00:26:48 +02:00
|
|
|
unsigned
|
|
|
|
DefaultIEW<Impl>::validInstsFromRename()
|
|
|
|
{
|
|
|
|
unsigned inst_count = 0;
|
|
|
|
|
|
|
|
for (int i=0; i<fromRename->size; i++) {
|
2006-06-14 19:12:41 +02:00
|
|
|
if (!fromRename->insts[i]->isSquashed())
|
2006-04-23 00:26:48 +02:00
|
|
|
inst_count++;
|
|
|
|
}
|
|
|
|
|
|
|
|
return inst_count;
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class Impl>
|
2004-08-20 20:54:07 +02:00
|
|
|
void
|
2009-05-26 18:23:13 +02:00
|
|
|
DefaultIEW<Impl>::skidInsert(ThreadID tid)
|
2006-04-23 00:26:48 +02:00
|
|
|
{
|
|
|
|
DynInstPtr inst = NULL;
|
Update to make multiple instruction issue and different latencies work.
Also change to ref counted DynInst.
SConscript:
Add branch predictor, BTB, load store queue, and storesets.
arch/isa_parser.py:
Specify the template parameter for AlphaDynInst
base/traceflags.py:
Add load store queue, store set, and mem dependence unit to the
list of trace flags.
cpu/base_dyn_inst.cc:
Change formating, add in debug statement.
cpu/base_dyn_inst.hh:
Change DynInst to be RefCounted, add flag to clear whether or not this
instruction can commit. This is likely to be removed in the future.
cpu/beta_cpu/alpha_dyn_inst.cc:
AlphaDynInst has been changed to be templated, so now this CC file
is just used to force instantiations of AlphaDynInst.
cpu/beta_cpu/alpha_dyn_inst.hh:
Changed AlphaDynInst to be templated on Impl. Removed some unnecessary
functions.
cpu/beta_cpu/alpha_full_cpu.cc:
AlphaFullCPU has been changed to be templated, so this CC file is now
just used to force instantation of AlphaFullCPU.
cpu/beta_cpu/alpha_full_cpu.hh:
Change AlphaFullCPU to be templated on Impl.
cpu/beta_cpu/alpha_impl.hh:
Update it to reflect AlphaDynInst and AlphaFullCPU being templated
on Impl. Also removed time buffers from here, as they are really
a part of the CPU and are thus in the CPU policy now.
cpu/beta_cpu/alpha_params.hh:
Make AlphaSimpleParams inherit from the BaseFullCPU so that it doesn't
need to specifically declare any parameters that are already in the
BaseFullCPU.
cpu/beta_cpu/comm.hh:
Changed the structure of the time buffer communication structs. Now
they include the size of the packet of instructions it is sending.
Added some parameters to the backwards communication struct, mainly
for squashing.
cpu/beta_cpu/commit.hh:
Update typenames to reflect change in location of time buffer structs.
Update DynInst to DynInstPtr (it is refcounted now).
cpu/beta_cpu/commit_impl.hh:
Formatting changes mainly. Also sends back proper information
on branch mispredicts so that the bpred unit can update itself.
Updated behavior for non-speculative instructions (stores, any
other non-spec instructions): once they reach the head of the ROB,
the ROB signals back to the IQ that it can go ahead and issue the
non-speculative instruction. The instruction itself is updated so that
commit won't try to commit it again until it is done executing.
cpu/beta_cpu/cpu_policy.hh:
Added branch prediction unit, mem dependence prediction unit, load
store queue. Moved time buffer structs from AlphaSimpleImpl to here.
cpu/beta_cpu/decode.hh:
Changed typedefs to reflect change in location of time buffer structs
and also the change from DynInst to ref counted DynInstPtr.
cpu/beta_cpu/decode_impl.hh:
Continues to buffer instructions even while unblocking now. Changed
how it loops through groups of instructions so it can properly block
during the middle of a group of instructions.
cpu/beta_cpu/fetch.hh:
Changed typedefs to reflect change in location of time buffer structs
and the change to ref counted DynInsts. Also added in branch
brediction unit.
cpu/beta_cpu/fetch_impl.hh:
Add in branch prediction. Changed how fetch checks inputs and its
current state to make for easier logic.
cpu/beta_cpu/free_list.cc:
Changed int regs and float regs to logically use one flat namespace.
Future change will be moving them to a single scoreboard to conserve
space.
cpu/beta_cpu/free_list.hh:
Mostly debugging statements. Might be removed for performance in future.
cpu/beta_cpu/full_cpu.cc:
Added in some debugging statements. Updated BaseFullCPU to take
a params object.
cpu/beta_cpu/full_cpu.hh:
Added params class within BaseCPU that other param classes will be
able to inherit from. Updated typedefs to reflect change in location
of time buffer structs and ref counted DynInst.
cpu/beta_cpu/iew.hh:
Updated typedefs to reflect change in location of time buffer structs
and use of ref counted DynInsts.
cpu/beta_cpu/iew_impl.hh:
Added in load store queue, updated iew to be able to execute non-
speculative instructions, instead of having them execute in commit.
cpu/beta_cpu/inst_queue.hh:
Updated change to ref counted DynInsts. Changed inst queue to hold
non-speculative instructions as well, which are issued only when
commit signals backwards that a nonspeculative instruction is at
the head of the ROB.
cpu/beta_cpu/inst_queue_impl.hh:
Updated to allow for non-speculative instructions to be in the inst
queue. Also added some debug functions.
cpu/beta_cpu/regfile.hh:
Added debugging statements, changed formatting.
cpu/beta_cpu/rename.hh:
Updated typedefs, added some functions to clean up code.
cpu/beta_cpu/rename_impl.hh:
Moved some code into functions to make it easier to read.
cpu/beta_cpu/rename_map.cc:
Changed int and float reg behavior to use a single flat namespace. In
the future, the rename maps can be combined to a single rename map to
save space.
cpu/beta_cpu/rename_map.hh:
Added destructor.
cpu/beta_cpu/rob.hh:
Updated it with change from DynInst to ref counted DynInst.
cpu/beta_cpu/rob_impl.hh:
Formatting, updated to use ref counted DynInst.
cpu/static_inst.hh:
Updated forward declaration for AlphaDynInst now that it is templated.
--HG--
extra : convert_revision : 1045f240ee9b6a4bd368e1806aca029ebbdc6dd3
2004-09-23 20:06:03 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
while (!insts[tid].empty()) {
|
|
|
|
inst = insts[tid].front();
|
2004-08-20 20:54:07 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
insts[tid].pop();
|
|
|
|
|
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
|
|
|
DPRINTF(Decode,"[tid:%i]: Inserting [sn:%lli] PC:%s into "
|
2006-04-23 00:26:48 +02:00
|
|
|
"dispatch skidBuffer %i\n",tid, inst->seqNum,
|
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
|
|
|
inst->pcState(),tid);
|
2006-04-23 00:26:48 +02:00
|
|
|
|
|
|
|
skidBuffer[tid].push(inst);
|
|
|
|
}
|
|
|
|
|
|
|
|
assert(skidBuffer[tid].size() <= skidBufferMax &&
|
|
|
|
"Skidbuffer Exceeded Max Size");
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class Impl>
|
|
|
|
int
|
|
|
|
DefaultIEW<Impl>::skidCount()
|
|
|
|
{
|
|
|
|
int max=0;
|
|
|
|
|
2009-05-26 18:23:13 +02:00
|
|
|
list<ThreadID>::iterator threads = activeThreads->begin();
|
|
|
|
list<ThreadID>::iterator end = activeThreads->end();
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2006-12-21 07:20:11 +01:00
|
|
|
while (threads != end) {
|
2009-05-26 18:23:13 +02:00
|
|
|
ThreadID tid = *threads++;
|
2006-12-21 07:20:11 +01:00
|
|
|
unsigned thread_count = skidBuffer[tid].size();
|
2006-04-23 00:26:48 +02:00
|
|
|
if (max < thread_count)
|
|
|
|
max = thread_count;
|
|
|
|
}
|
|
|
|
|
|
|
|
return max;
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class Impl>
|
|
|
|
bool
|
|
|
|
DefaultIEW<Impl>::skidsEmpty()
|
|
|
|
{
|
2009-05-26 18:23:13 +02:00
|
|
|
list<ThreadID>::iterator threads = activeThreads->begin();
|
|
|
|
list<ThreadID>::iterator end = activeThreads->end();
|
2006-12-21 07:20:11 +01:00
|
|
|
|
|
|
|
while (threads != end) {
|
2009-05-26 18:23:13 +02:00
|
|
|
ThreadID tid = *threads++;
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2006-12-21 07:20:11 +01:00
|
|
|
if (!skidBuffer[tid].empty())
|
2006-04-23 00:26:48 +02:00
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
void
|
|
|
|
DefaultIEW<Impl>::updateStatus()
|
|
|
|
{
|
|
|
|
bool any_unblocking = false;
|
|
|
|
|
2009-05-26 18:23:13 +02:00
|
|
|
list<ThreadID>::iterator threads = activeThreads->begin();
|
|
|
|
list<ThreadID>::iterator end = activeThreads->end();
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2006-12-21 07:20:11 +01:00
|
|
|
while (threads != end) {
|
2009-05-26 18:23:13 +02:00
|
|
|
ThreadID tid = *threads++;
|
2006-04-23 00:26:48 +02:00
|
|
|
|
|
|
|
if (dispatchStatus[tid] == Unblocking) {
|
|
|
|
any_unblocking = true;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// If there are no ready instructions waiting to be scheduled by the IQ,
|
|
|
|
// and there's no stores waiting to write back, and dispatch is not
|
|
|
|
// unblocking, then there is no internal activity for the IEW stage.
|
2011-02-07 07:14:17 +01:00
|
|
|
instQueue.intInstQueueReads++;
|
2006-04-23 00:26:48 +02:00
|
|
|
if (_status == Active && !instQueue.hasReadyInsts() &&
|
|
|
|
!ldstQueue.willWB() && !any_unblocking) {
|
|
|
|
DPRINTF(IEW, "IEW switching to idle\n");
|
|
|
|
|
|
|
|
deactivateStage();
|
|
|
|
|
|
|
|
_status = Inactive;
|
|
|
|
} else if (_status == Inactive && (instQueue.hasReadyInsts() ||
|
|
|
|
ldstQueue.willWB() ||
|
|
|
|
any_unblocking)) {
|
|
|
|
// Otherwise there is internal activity. Set to active.
|
|
|
|
DPRINTF(IEW, "IEW switching to active\n");
|
|
|
|
|
|
|
|
activateStage();
|
|
|
|
|
|
|
|
_status = Active;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
void
|
|
|
|
DefaultIEW<Impl>::resetEntries()
|
|
|
|
{
|
|
|
|
instQueue.resetEntries();
|
|
|
|
ldstQueue.resetEntries();
|
|
|
|
}
|
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
void
|
2009-05-26 18:23:13 +02:00
|
|
|
DefaultIEW<Impl>::readStallSignals(ThreadID tid)
|
2006-04-23 00:26:48 +02:00
|
|
|
{
|
|
|
|
if (fromCommit->commitBlock[tid]) {
|
|
|
|
stalls[tid].commit = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (fromCommit->commitUnblock[tid]) {
|
|
|
|
assert(stalls[tid].commit);
|
|
|
|
stalls[tid].commit = false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
bool
|
2009-05-26 18:23:13 +02:00
|
|
|
DefaultIEW<Impl>::checkStall(ThreadID tid)
|
2006-04-23 00:26:48 +02:00
|
|
|
{
|
|
|
|
bool ret_val(false);
|
|
|
|
|
|
|
|
if (stalls[tid].commit) {
|
|
|
|
DPRINTF(IEW,"[tid:%i]: Stall from Commit stage detected.\n",tid);
|
|
|
|
ret_val = true;
|
|
|
|
} else if (instQueue.isFull(tid)) {
|
|
|
|
DPRINTF(IEW,"[tid:%i]: Stall: IQ is full.\n",tid);
|
|
|
|
ret_val = true;
|
|
|
|
} else if (ldstQueue.isFull(tid)) {
|
|
|
|
DPRINTF(IEW,"[tid:%i]: Stall: LSQ is full\n",tid);
|
|
|
|
|
|
|
|
if (ldstQueue.numLoads(tid) > 0 ) {
|
|
|
|
|
|
|
|
DPRINTF(IEW,"[tid:%i]: LSQ oldest load: [sn:%i] \n",
|
|
|
|
tid,ldstQueue.getLoadHeadSeqNum(tid));
|
|
|
|
}
|
|
|
|
|
|
|
|
if (ldstQueue.numStores(tid) > 0) {
|
|
|
|
|
|
|
|
DPRINTF(IEW,"[tid:%i]: LSQ oldest store: [sn:%i] \n",
|
|
|
|
tid,ldstQueue.getStoreHeadSeqNum(tid));
|
|
|
|
}
|
|
|
|
|
|
|
|
ret_val = true;
|
|
|
|
} else if (ldstQueue.isStalled(tid)) {
|
|
|
|
DPRINTF(IEW,"[tid:%i]: Stall: LSQ stall detected.\n",tid);
|
|
|
|
ret_val = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret_val;
|
|
|
|
}
|
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
void
|
2009-05-26 18:23:13 +02:00
|
|
|
DefaultIEW<Impl>::checkSignalsAndUpdate(ThreadID tid)
|
2006-04-23 00:26:48 +02:00
|
|
|
{
|
|
|
|
// Check if there's a squash signal, squash if there is
|
|
|
|
// Check stall signals, block if there is.
|
|
|
|
// If status was Blocked
|
|
|
|
// if so then go to unblocking
|
|
|
|
// If status was Squashing
|
|
|
|
// check if squashing is not high. Switch to running this cycle.
|
|
|
|
|
|
|
|
readStallSignals(tid);
|
|
|
|
|
|
|
|
if (fromCommit->commitInfo[tid].squash) {
|
|
|
|
squash(tid);
|
|
|
|
|
|
|
|
if (dispatchStatus[tid] == Blocked ||
|
|
|
|
dispatchStatus[tid] == Unblocking) {
|
|
|
|
toRename->iewUnblock[tid] = true;
|
|
|
|
wroteToTimeBuffer = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
dispatchStatus[tid] = Squashing;
|
|
|
|
fetchRedirect[tid] = false;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (fromCommit->commitInfo[tid].robSquashing) {
|
2006-06-13 01:04:42 +02:00
|
|
|
DPRINTF(IEW, "[tid:%i]: ROB is still squashing.\n", tid);
|
2006-04-23 00:26:48 +02:00
|
|
|
|
|
|
|
dispatchStatus[tid] = Squashing;
|
2006-06-13 01:04:42 +02:00
|
|
|
emptyRenameInsts(tid);
|
|
|
|
wroteToTimeBuffer = true;
|
2006-04-23 00:26:48 +02:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (checkStall(tid)) {
|
|
|
|
block(tid);
|
|
|
|
dispatchStatus[tid] = Blocked;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (dispatchStatus[tid] == Blocked) {
|
|
|
|
// Status from previous cycle was blocked, but there are no more stall
|
|
|
|
// conditions. Switch over to unblocking.
|
|
|
|
DPRINTF(IEW, "[tid:%i]: Done blocking, switching to unblocking.\n",
|
|
|
|
tid);
|
|
|
|
|
|
|
|
dispatchStatus[tid] = Unblocking;
|
|
|
|
|
|
|
|
unblock(tid);
|
|
|
|
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (dispatchStatus[tid] == Squashing) {
|
|
|
|
// Switch status to running if rename isn't being told to block or
|
|
|
|
// squash this cycle.
|
|
|
|
DPRINTF(IEW, "[tid:%i]: Done squashing, switching to running.\n",
|
|
|
|
tid);
|
|
|
|
|
|
|
|
dispatchStatus[tid] = Running;
|
|
|
|
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
void
|
|
|
|
DefaultIEW<Impl>::sortInsts()
|
|
|
|
{
|
|
|
|
int insts_from_rename = fromRename->size;
|
2006-05-19 21:44:03 +02:00
|
|
|
#ifdef DEBUG
|
2009-05-26 18:23:13 +02:00
|
|
|
for (ThreadID tid = 0; tid < numThreads; tid++)
|
|
|
|
assert(insts[tid].empty());
|
2006-05-19 21:44:03 +02:00
|
|
|
#endif
|
2006-04-23 00:26:48 +02:00
|
|
|
for (int i = 0; i < insts_from_rename; ++i) {
|
|
|
|
insts[fromRename->insts[i]->threadNumber].push(fromRename->insts[i]);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2006-06-13 01:04:42 +02:00
|
|
|
template <class Impl>
|
|
|
|
void
|
2009-05-26 18:23:13 +02:00
|
|
|
DefaultIEW<Impl>::emptyRenameInsts(ThreadID tid)
|
2006-06-13 01:04:42 +02:00
|
|
|
{
|
2007-04-13 15:59:31 +02:00
|
|
|
DPRINTF(IEW, "[tid:%i]: Removing incoming rename instructions\n", tid);
|
2006-07-23 19:39:42 +02:00
|
|
|
|
2006-06-13 01:04:42 +02:00
|
|
|
while (!insts[tid].empty()) {
|
2006-07-23 19:39:42 +02:00
|
|
|
|
2006-06-13 01:04:42 +02:00
|
|
|
if (insts[tid].front()->isLoad() ||
|
|
|
|
insts[tid].front()->isStore() ) {
|
|
|
|
toRename->iewInfo[tid].dispatchedToLSQ++;
|
2006-08-02 18:05:34 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
toRename->iewInfo[tid].dispatched++;
|
|
|
|
|
|
|
|
insts[tid].pop();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
template <class Impl>
|
|
|
|
void
|
|
|
|
DefaultIEW<Impl>::wakeCPU()
|
|
|
|
{
|
|
|
|
cpu->wakeCPU();
|
|
|
|
}
|
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
void
|
|
|
|
DefaultIEW<Impl>::activityThisCycle()
|
|
|
|
{
|
|
|
|
DPRINTF(Activity, "Activity this cycle.\n");
|
|
|
|
cpu->activityThisCycle();
|
|
|
|
}
|
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
inline void
|
|
|
|
DefaultIEW<Impl>::activateStage()
|
|
|
|
{
|
|
|
|
DPRINTF(Activity, "Activating stage.\n");
|
2006-06-16 23:08:47 +02:00
|
|
|
cpu->activateStage(O3CPU::IEWIdx);
|
2006-04-23 00:26:48 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
inline void
|
|
|
|
DefaultIEW<Impl>::deactivateStage()
|
|
|
|
{
|
|
|
|
DPRINTF(Activity, "Deactivating stage.\n");
|
2006-06-16 23:08:47 +02:00
|
|
|
cpu->deactivateStage(O3CPU::IEWIdx);
|
2006-04-23 00:26:48 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
template<class Impl>
|
|
|
|
void
|
2009-05-26 18:23:13 +02:00
|
|
|
DefaultIEW<Impl>::dispatch(ThreadID tid)
|
2006-04-23 00:26:48 +02:00
|
|
|
{
|
|
|
|
// If status is Running or idle,
|
|
|
|
// call dispatchInsts()
|
|
|
|
// If status is Unblocking,
|
|
|
|
// buffer any instructions coming from rename
|
|
|
|
// continue trying to empty skid buffer
|
|
|
|
// check if stall conditions have passed
|
|
|
|
|
|
|
|
if (dispatchStatus[tid] == Blocked) {
|
|
|
|
++iewBlockCycles;
|
|
|
|
|
|
|
|
} else if (dispatchStatus[tid] == Squashing) {
|
|
|
|
++iewSquashCycles;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Dispatch should try to dispatch as many instructions as its bandwidth
|
|
|
|
// will allow, as long as it is not currently blocked.
|
|
|
|
if (dispatchStatus[tid] == Running ||
|
|
|
|
dispatchStatus[tid] == Idle) {
|
|
|
|
DPRINTF(IEW, "[tid:%i] Not blocked, so attempting to run "
|
|
|
|
"dispatch.\n", tid);
|
|
|
|
|
|
|
|
dispatchInsts(tid);
|
|
|
|
} else if (dispatchStatus[tid] == Unblocking) {
|
|
|
|
// Make sure that the skid buffer has something in it if the
|
|
|
|
// status is unblocking.
|
|
|
|
assert(!skidsEmpty());
|
|
|
|
|
|
|
|
// If the status was unblocking, then instructions from the skid
|
|
|
|
// buffer were used. Remove those instructions and handle
|
|
|
|
// the rest of unblocking.
|
|
|
|
dispatchInsts(tid);
|
|
|
|
|
|
|
|
++iewUnblockCycles;
|
|
|
|
|
2007-11-06 21:51:08 +01:00
|
|
|
if (validInstsFromRename()) {
|
2006-04-23 00:26:48 +02:00
|
|
|
// Add the current inputs to the skid buffer so they can be
|
|
|
|
// reprocessed when this stage unblocks.
|
|
|
|
skidInsert(tid);
|
|
|
|
}
|
|
|
|
|
|
|
|
unblock(tid);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
void
|
2009-05-26 18:23:13 +02:00
|
|
|
DefaultIEW<Impl>::dispatchInsts(ThreadID tid)
|
2006-04-23 00:26:48 +02:00
|
|
|
{
|
|
|
|
// Obtain instructions from skid buffer if unblocking, or queue from rename
|
|
|
|
// otherwise.
|
|
|
|
std::queue<DynInstPtr> &insts_to_dispatch =
|
|
|
|
dispatchStatus[tid] == Unblocking ?
|
|
|
|
skidBuffer[tid] : insts[tid];
|
|
|
|
|
|
|
|
int insts_to_add = insts_to_dispatch.size();
|
|
|
|
|
|
|
|
DynInstPtr inst;
|
|
|
|
bool add_to_iq = false;
|
|
|
|
int dis_num_inst = 0;
|
|
|
|
|
|
|
|
// Loop through the instructions, putting them in the instruction
|
|
|
|
// queue.
|
|
|
|
for ( ; dis_num_inst < insts_to_add &&
|
2006-07-05 21:51:36 +02:00
|
|
|
dis_num_inst < dispatchWidth;
|
2006-04-23 00:26:48 +02:00
|
|
|
++dis_num_inst)
|
|
|
|
{
|
|
|
|
inst = insts_to_dispatch.front();
|
|
|
|
|
|
|
|
if (dispatchStatus[tid] == Unblocking) {
|
|
|
|
DPRINTF(IEW, "[tid:%i]: Issue: Examining instruction from skid "
|
|
|
|
"buffer\n", tid);
|
|
|
|
}
|
|
|
|
|
|
|
|
// Make sure there's a valid instruction there.
|
|
|
|
assert(inst);
|
2004-08-20 20:54:07 +02:00
|
|
|
|
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
|
|
|
DPRINTF(IEW, "[tid:%i]: Issue: Adding PC %s [sn:%lli] [tid:%i] to "
|
2006-04-23 00:26:48 +02:00
|
|
|
"IQ.\n",
|
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
|
|
|
tid, inst->pcState(), inst->seqNum, inst->threadNumber);
|
2006-04-23 00:26:48 +02:00
|
|
|
|
|
|
|
// Be sure to mark these instructions as ready so that the
|
|
|
|
// commit stage can go ahead and execute them, and mark
|
|
|
|
// them as issued so the IQ doesn't reprocess them.
|
|
|
|
|
|
|
|
// Check for squashed instructions.
|
|
|
|
if (inst->isSquashed()) {
|
|
|
|
DPRINTF(IEW, "[tid:%i]: Issue: Squashed instruction encountered, "
|
|
|
|
"not adding to IQ.\n", tid);
|
|
|
|
|
|
|
|
++iewDispSquashedInsts;
|
|
|
|
|
|
|
|
insts_to_dispatch.pop();
|
|
|
|
|
|
|
|
//Tell Rename That An Instruction has been processed
|
|
|
|
if (inst->isLoad() || inst->isStore()) {
|
|
|
|
toRename->iewInfo[tid].dispatchedToLSQ++;
|
|
|
|
}
|
|
|
|
toRename->iewInfo[tid].dispatched++;
|
|
|
|
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Check for full conditions.
|
|
|
|
if (instQueue.isFull(tid)) {
|
|
|
|
DPRINTF(IEW, "[tid:%i]: Issue: IQ has become full.\n", tid);
|
|
|
|
|
|
|
|
// Call function to start blocking.
|
|
|
|
block(tid);
|
|
|
|
|
|
|
|
// Set unblock to false. Special case where we are using
|
|
|
|
// skidbuffer (unblocking) instructions but then we still
|
|
|
|
// get full in the IQ.
|
|
|
|
toRename->iewUnblock[tid] = false;
|
|
|
|
|
|
|
|
++iewIQFullEvents;
|
|
|
|
break;
|
|
|
|
} else if (ldstQueue.isFull(tid)) {
|
|
|
|
DPRINTF(IEW, "[tid:%i]: Issue: LSQ has become full.\n",tid);
|
|
|
|
|
|
|
|
// Call function to start blocking.
|
|
|
|
block(tid);
|
|
|
|
|
|
|
|
// Set unblock to false. Special case where we are using
|
|
|
|
// skidbuffer (unblocking) instructions but then we still
|
|
|
|
// get full in the IQ.
|
|
|
|
toRename->iewUnblock[tid] = false;
|
|
|
|
|
|
|
|
++iewLSQFullEvents;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Otherwise issue the instruction just fine.
|
|
|
|
if (inst->isLoad()) {
|
|
|
|
DPRINTF(IEW, "[tid:%i]: Issue: Memory instruction "
|
|
|
|
"encountered, adding to LSQ.\n", tid);
|
|
|
|
|
|
|
|
// Reserve a spot in the load store queue for this
|
|
|
|
// memory access.
|
|
|
|
ldstQueue.insertLoad(inst);
|
|
|
|
|
|
|
|
++iewDispLoadInsts;
|
|
|
|
|
|
|
|
add_to_iq = true;
|
|
|
|
|
|
|
|
toRename->iewInfo[tid].dispatchedToLSQ++;
|
|
|
|
} else if (inst->isStore()) {
|
|
|
|
DPRINTF(IEW, "[tid:%i]: Issue: Memory instruction "
|
|
|
|
"encountered, adding to LSQ.\n", tid);
|
|
|
|
|
|
|
|
ldstQueue.insertStore(inst);
|
|
|
|
|
|
|
|
++iewDispStoreInsts;
|
|
|
|
|
2006-05-23 20:38:16 +02:00
|
|
|
if (inst->isStoreConditional()) {
|
|
|
|
// Store conditionals need to be set as "canCommit()"
|
|
|
|
// so that commit can process them when they reach the
|
|
|
|
// head of commit.
|
2006-05-31 17:45:02 +02:00
|
|
|
// @todo: This is somewhat specific to Alpha.
|
2006-04-23 00:26:48 +02:00
|
|
|
inst->setCanCommit();
|
Update to make multiple instruction issue and different latencies work.
Also change to ref counted DynInst.
SConscript:
Add branch predictor, BTB, load store queue, and storesets.
arch/isa_parser.py:
Specify the template parameter for AlphaDynInst
base/traceflags.py:
Add load store queue, store set, and mem dependence unit to the
list of trace flags.
cpu/base_dyn_inst.cc:
Change formating, add in debug statement.
cpu/base_dyn_inst.hh:
Change DynInst to be RefCounted, add flag to clear whether or not this
instruction can commit. This is likely to be removed in the future.
cpu/beta_cpu/alpha_dyn_inst.cc:
AlphaDynInst has been changed to be templated, so now this CC file
is just used to force instantiations of AlphaDynInst.
cpu/beta_cpu/alpha_dyn_inst.hh:
Changed AlphaDynInst to be templated on Impl. Removed some unnecessary
functions.
cpu/beta_cpu/alpha_full_cpu.cc:
AlphaFullCPU has been changed to be templated, so this CC file is now
just used to force instantation of AlphaFullCPU.
cpu/beta_cpu/alpha_full_cpu.hh:
Change AlphaFullCPU to be templated on Impl.
cpu/beta_cpu/alpha_impl.hh:
Update it to reflect AlphaDynInst and AlphaFullCPU being templated
on Impl. Also removed time buffers from here, as they are really
a part of the CPU and are thus in the CPU policy now.
cpu/beta_cpu/alpha_params.hh:
Make AlphaSimpleParams inherit from the BaseFullCPU so that it doesn't
need to specifically declare any parameters that are already in the
BaseFullCPU.
cpu/beta_cpu/comm.hh:
Changed the structure of the time buffer communication structs. Now
they include the size of the packet of instructions it is sending.
Added some parameters to the backwards communication struct, mainly
for squashing.
cpu/beta_cpu/commit.hh:
Update typenames to reflect change in location of time buffer structs.
Update DynInst to DynInstPtr (it is refcounted now).
cpu/beta_cpu/commit_impl.hh:
Formatting changes mainly. Also sends back proper information
on branch mispredicts so that the bpred unit can update itself.
Updated behavior for non-speculative instructions (stores, any
other non-spec instructions): once they reach the head of the ROB,
the ROB signals back to the IQ that it can go ahead and issue the
non-speculative instruction. The instruction itself is updated so that
commit won't try to commit it again until it is done executing.
cpu/beta_cpu/cpu_policy.hh:
Added branch prediction unit, mem dependence prediction unit, load
store queue. Moved time buffer structs from AlphaSimpleImpl to here.
cpu/beta_cpu/decode.hh:
Changed typedefs to reflect change in location of time buffer structs
and also the change from DynInst to ref counted DynInstPtr.
cpu/beta_cpu/decode_impl.hh:
Continues to buffer instructions even while unblocking now. Changed
how it loops through groups of instructions so it can properly block
during the middle of a group of instructions.
cpu/beta_cpu/fetch.hh:
Changed typedefs to reflect change in location of time buffer structs
and the change to ref counted DynInsts. Also added in branch
brediction unit.
cpu/beta_cpu/fetch_impl.hh:
Add in branch prediction. Changed how fetch checks inputs and its
current state to make for easier logic.
cpu/beta_cpu/free_list.cc:
Changed int regs and float regs to logically use one flat namespace.
Future change will be moving them to a single scoreboard to conserve
space.
cpu/beta_cpu/free_list.hh:
Mostly debugging statements. Might be removed for performance in future.
cpu/beta_cpu/full_cpu.cc:
Added in some debugging statements. Updated BaseFullCPU to take
a params object.
cpu/beta_cpu/full_cpu.hh:
Added params class within BaseCPU that other param classes will be
able to inherit from. Updated typedefs to reflect change in location
of time buffer structs and ref counted DynInst.
cpu/beta_cpu/iew.hh:
Updated typedefs to reflect change in location of time buffer structs
and use of ref counted DynInsts.
cpu/beta_cpu/iew_impl.hh:
Added in load store queue, updated iew to be able to execute non-
speculative instructions, instead of having them execute in commit.
cpu/beta_cpu/inst_queue.hh:
Updated change to ref counted DynInsts. Changed inst queue to hold
non-speculative instructions as well, which are issued only when
commit signals backwards that a nonspeculative instruction is at
the head of the ROB.
cpu/beta_cpu/inst_queue_impl.hh:
Updated to allow for non-speculative instructions to be in the inst
queue. Also added some debug functions.
cpu/beta_cpu/regfile.hh:
Added debugging statements, changed formatting.
cpu/beta_cpu/rename.hh:
Updated typedefs, added some functions to clean up code.
cpu/beta_cpu/rename_impl.hh:
Moved some code into functions to make it easier to read.
cpu/beta_cpu/rename_map.cc:
Changed int and float reg behavior to use a single flat namespace. In
the future, the rename maps can be combined to a single rename map to
save space.
cpu/beta_cpu/rename_map.hh:
Added destructor.
cpu/beta_cpu/rob.hh:
Updated it with change from DynInst to ref counted DynInst.
cpu/beta_cpu/rob_impl.hh:
Formatting, updated to use ref counted DynInst.
cpu/static_inst.hh:
Updated forward declaration for AlphaDynInst now that it is templated.
--HG--
extra : convert_revision : 1045f240ee9b6a4bd368e1806aca029ebbdc6dd3
2004-09-23 20:06:03 +02:00
|
|
|
instQueue.insertNonSpec(inst);
|
2006-04-23 00:26:48 +02:00
|
|
|
add_to_iq = false;
|
Update to make multiple instruction issue and different latencies work.
Also change to ref counted DynInst.
SConscript:
Add branch predictor, BTB, load store queue, and storesets.
arch/isa_parser.py:
Specify the template parameter for AlphaDynInst
base/traceflags.py:
Add load store queue, store set, and mem dependence unit to the
list of trace flags.
cpu/base_dyn_inst.cc:
Change formating, add in debug statement.
cpu/base_dyn_inst.hh:
Change DynInst to be RefCounted, add flag to clear whether or not this
instruction can commit. This is likely to be removed in the future.
cpu/beta_cpu/alpha_dyn_inst.cc:
AlphaDynInst has been changed to be templated, so now this CC file
is just used to force instantiations of AlphaDynInst.
cpu/beta_cpu/alpha_dyn_inst.hh:
Changed AlphaDynInst to be templated on Impl. Removed some unnecessary
functions.
cpu/beta_cpu/alpha_full_cpu.cc:
AlphaFullCPU has been changed to be templated, so this CC file is now
just used to force instantation of AlphaFullCPU.
cpu/beta_cpu/alpha_full_cpu.hh:
Change AlphaFullCPU to be templated on Impl.
cpu/beta_cpu/alpha_impl.hh:
Update it to reflect AlphaDynInst and AlphaFullCPU being templated
on Impl. Also removed time buffers from here, as they are really
a part of the CPU and are thus in the CPU policy now.
cpu/beta_cpu/alpha_params.hh:
Make AlphaSimpleParams inherit from the BaseFullCPU so that it doesn't
need to specifically declare any parameters that are already in the
BaseFullCPU.
cpu/beta_cpu/comm.hh:
Changed the structure of the time buffer communication structs. Now
they include the size of the packet of instructions it is sending.
Added some parameters to the backwards communication struct, mainly
for squashing.
cpu/beta_cpu/commit.hh:
Update typenames to reflect change in location of time buffer structs.
Update DynInst to DynInstPtr (it is refcounted now).
cpu/beta_cpu/commit_impl.hh:
Formatting changes mainly. Also sends back proper information
on branch mispredicts so that the bpred unit can update itself.
Updated behavior for non-speculative instructions (stores, any
other non-spec instructions): once they reach the head of the ROB,
the ROB signals back to the IQ that it can go ahead and issue the
non-speculative instruction. The instruction itself is updated so that
commit won't try to commit it again until it is done executing.
cpu/beta_cpu/cpu_policy.hh:
Added branch prediction unit, mem dependence prediction unit, load
store queue. Moved time buffer structs from AlphaSimpleImpl to here.
cpu/beta_cpu/decode.hh:
Changed typedefs to reflect change in location of time buffer structs
and also the change from DynInst to ref counted DynInstPtr.
cpu/beta_cpu/decode_impl.hh:
Continues to buffer instructions even while unblocking now. Changed
how it loops through groups of instructions so it can properly block
during the middle of a group of instructions.
cpu/beta_cpu/fetch.hh:
Changed typedefs to reflect change in location of time buffer structs
and the change to ref counted DynInsts. Also added in branch
brediction unit.
cpu/beta_cpu/fetch_impl.hh:
Add in branch prediction. Changed how fetch checks inputs and its
current state to make for easier logic.
cpu/beta_cpu/free_list.cc:
Changed int regs and float regs to logically use one flat namespace.
Future change will be moving them to a single scoreboard to conserve
space.
cpu/beta_cpu/free_list.hh:
Mostly debugging statements. Might be removed for performance in future.
cpu/beta_cpu/full_cpu.cc:
Added in some debugging statements. Updated BaseFullCPU to take
a params object.
cpu/beta_cpu/full_cpu.hh:
Added params class within BaseCPU that other param classes will be
able to inherit from. Updated typedefs to reflect change in location
of time buffer structs and ref counted DynInst.
cpu/beta_cpu/iew.hh:
Updated typedefs to reflect change in location of time buffer structs
and use of ref counted DynInsts.
cpu/beta_cpu/iew_impl.hh:
Added in load store queue, updated iew to be able to execute non-
speculative instructions, instead of having them execute in commit.
cpu/beta_cpu/inst_queue.hh:
Updated change to ref counted DynInsts. Changed inst queue to hold
non-speculative instructions as well, which are issued only when
commit signals backwards that a nonspeculative instruction is at
the head of the ROB.
cpu/beta_cpu/inst_queue_impl.hh:
Updated to allow for non-speculative instructions to be in the inst
queue. Also added some debug functions.
cpu/beta_cpu/regfile.hh:
Added debugging statements, changed formatting.
cpu/beta_cpu/rename.hh:
Updated typedefs, added some functions to clean up code.
cpu/beta_cpu/rename_impl.hh:
Moved some code into functions to make it easier to read.
cpu/beta_cpu/rename_map.cc:
Changed int and float reg behavior to use a single flat namespace. In
the future, the rename maps can be combined to a single rename map to
save space.
cpu/beta_cpu/rename_map.hh:
Added destructor.
cpu/beta_cpu/rob.hh:
Updated it with change from DynInst to ref counted DynInst.
cpu/beta_cpu/rob_impl.hh:
Formatting, updated to use ref counted DynInst.
cpu/static_inst.hh:
Updated forward declaration for AlphaDynInst now that it is templated.
--HG--
extra : convert_revision : 1045f240ee9b6a4bd368e1806aca029ebbdc6dd3
2004-09-23 20:06:03 +02:00
|
|
|
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
++iewDispNonSpecInsts;
|
2006-04-23 00:26:48 +02:00
|
|
|
} else {
|
|
|
|
add_to_iq = true;
|
|
|
|
}
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
toRename->iewInfo[tid].dispatchedToLSQ++;
|
|
|
|
} else if (inst->isMemBarrier() || inst->isWriteBarrier()) {
|
2006-05-19 21:44:03 +02:00
|
|
|
// Same as non-speculative stores.
|
2006-04-23 00:26:48 +02:00
|
|
|
inst->setCanCommit();
|
|
|
|
instQueue.insertBarrier(inst);
|
|
|
|
add_to_iq = false;
|
|
|
|
} else if (inst->isNop()) {
|
|
|
|
DPRINTF(IEW, "[tid:%i]: Issue: Nop instruction encountered, "
|
|
|
|
"skipping.\n", tid);
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
inst->setIssued();
|
|
|
|
inst->setExecuted();
|
|
|
|
inst->setCanCommit();
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-05-19 21:44:03 +02:00
|
|
|
instQueue.recordProducer(inst);
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2006-06-14 04:35:05 +02:00
|
|
|
iewExecutedNop[tid]++;
|
2006-04-24 23:06:00 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
add_to_iq = false;
|
|
|
|
} else if (inst->isExecuted()) {
|
|
|
|
assert(0 && "Instruction shouldn't be executed.\n");
|
|
|
|
DPRINTF(IEW, "Issue: Executed branch encountered, "
|
|
|
|
"skipping.\n");
|
2004-08-20 20:54:07 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
inst->setIssued();
|
|
|
|
inst->setCanCommit();
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-05-19 21:44:03 +02:00
|
|
|
instQueue.recordProducer(inst);
|
2006-04-23 00:26:48 +02:00
|
|
|
|
|
|
|
add_to_iq = false;
|
|
|
|
} else {
|
|
|
|
add_to_iq = true;
|
2004-08-20 20:54:07 +02:00
|
|
|
}
|
2007-03-23 16:40:53 +01:00
|
|
|
if (inst->isNonSpeculative()) {
|
|
|
|
DPRINTF(IEW, "[tid:%i]: Issue: Nonspeculative instruction "
|
|
|
|
"encountered, skipping.\n", tid);
|
|
|
|
|
|
|
|
// Same as non-speculative stores.
|
|
|
|
inst->setCanCommit();
|
|
|
|
|
|
|
|
// Specifically insert it as nonspeculative.
|
|
|
|
instQueue.insertNonSpec(inst);
|
|
|
|
|
|
|
|
++iewDispNonSpecInsts;
|
|
|
|
|
|
|
|
add_to_iq = false;
|
|
|
|
}
|
2006-04-23 00:26:48 +02:00
|
|
|
|
|
|
|
// If the instruction queue is not full, then add the
|
|
|
|
// instruction.
|
|
|
|
if (add_to_iq) {
|
|
|
|
instQueue.insert(inst);
|
|
|
|
}
|
|
|
|
|
|
|
|
insts_to_dispatch.pop();
|
|
|
|
|
|
|
|
toRename->iewInfo[tid].dispatched++;
|
|
|
|
|
|
|
|
++iewDispatchedInsts;
|
2011-07-15 18:53:35 +02:00
|
|
|
|
|
|
|
#if TRACING_ON
|
|
|
|
inst->dispatchTick = curTick();
|
|
|
|
#endif
|
2006-04-23 00:26:48 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
if (!insts_to_dispatch.empty()) {
|
2006-07-23 19:39:42 +02:00
|
|
|
DPRINTF(IEW,"[tid:%i]: Issue: Bandwidth Full. Blocking.\n", tid);
|
2006-04-23 00:26:48 +02:00
|
|
|
block(tid);
|
|
|
|
toRename->iewUnblock[tid] = false;
|
2004-08-20 20:54:07 +02:00
|
|
|
}
|
2006-04-23 00:26:48 +02:00
|
|
|
|
|
|
|
if (dispatchStatus[tid] == Idle && dis_num_inst) {
|
|
|
|
dispatchStatus[tid] = Running;
|
|
|
|
|
|
|
|
updatedQueues = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
dis_num_inst = 0;
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
}
|
2004-08-20 20:54:07 +02:00
|
|
|
|
2005-05-03 16:56:47 +02:00
|
|
|
template <class Impl>
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
void
|
2006-04-23 00:26:48 +02:00
|
|
|
DefaultIEW<Impl>::printAvailableInsts()
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
{
|
2006-04-23 00:26:48 +02:00
|
|
|
int inst = 0;
|
|
|
|
|
2006-08-15 11:07:15 +02:00
|
|
|
std::cout << "Available Instructions: ";
|
2006-04-23 00:26:48 +02:00
|
|
|
|
|
|
|
while (fromIssue->insts[inst]) {
|
|
|
|
|
2006-08-15 11:07:15 +02:00
|
|
|
if (inst%3==0) std::cout << "\n\t";
|
2006-04-23 00:26:48 +02:00
|
|
|
|
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
|
|
|
std::cout << "PC: " << fromIssue->insts[inst]->pcState()
|
2006-04-23 00:26:48 +02:00
|
|
|
<< " TN: " << fromIssue->insts[inst]->threadNumber
|
|
|
|
<< " SN: " << fromIssue->insts[inst]->seqNum << " | ";
|
2004-08-20 20:54:07 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
inst++;
|
2004-08-20 20:54:07 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
}
|
|
|
|
|
2006-08-15 11:07:15 +02:00
|
|
|
std::cout << "\n";
|
2006-04-23 00:26:48 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
void
|
|
|
|
DefaultIEW<Impl>::executeInsts()
|
|
|
|
{
|
|
|
|
wbNumInst = 0;
|
|
|
|
wbCycle = 0;
|
|
|
|
|
2009-05-26 18:23:13 +02:00
|
|
|
list<ThreadID>::iterator threads = activeThreads->begin();
|
|
|
|
list<ThreadID>::iterator end = activeThreads->end();
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2006-12-21 07:20:11 +01:00
|
|
|
while (threads != end) {
|
2009-05-26 18:23:13 +02:00
|
|
|
ThreadID tid = *threads++;
|
2006-04-23 00:26:48 +02:00
|
|
|
fetchRedirect[tid] = false;
|
|
|
|
}
|
|
|
|
|
2006-06-09 22:28:17 +02:00
|
|
|
// Uncomment this if you want to see all available instructions.
|
2010-08-23 18:18:41 +02:00
|
|
|
// @todo This doesn't actually work anymore, we should fix it.
|
2006-06-09 22:28:17 +02:00
|
|
|
// printAvailableInsts();
|
2004-08-20 20:54:07 +02:00
|
|
|
|
|
|
|
// Execute/writeback any instructions that are available.
|
2006-05-23 23:03:43 +02:00
|
|
|
int insts_to_execute = fromIssue->size;
|
2006-04-23 00:26:48 +02:00
|
|
|
int inst_num = 0;
|
2006-05-23 23:03:43 +02:00
|
|
|
for (; inst_num < insts_to_execute;
|
2006-05-19 21:44:03 +02:00
|
|
|
++inst_num) {
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
DPRINTF(IEW, "Execute: Executing instructions from IQ.\n");
|
2004-08-20 20:54:07 +02:00
|
|
|
|
2006-05-23 23:03:43 +02:00
|
|
|
DynInstPtr inst = instQueue.getInstToExecute();
|
2004-08-20 20:54:07 +02:00
|
|
|
|
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
|
|
|
DPRINTF(IEW, "Execute: Processing PC %s, [tid:%i] [sn:%i].\n",
|
|
|
|
inst->pcState(), inst->threadNumber,inst->seqNum);
|
2004-08-20 20:54:07 +02:00
|
|
|
|
|
|
|
// Check if the instruction is squashed; if so then skip it
|
|
|
|
if (inst->isSquashed()) {
|
2011-05-23 17:40:19 +02:00
|
|
|
DPRINTF(IEW, "Execute: Instruction was squashed. PC: %s, [tid:%i]"
|
|
|
|
" [sn:%i]\n", inst->pcState(), inst->threadNumber,
|
|
|
|
inst->seqNum);
|
Update to make multiple instruction issue and different latencies work.
Also change to ref counted DynInst.
SConscript:
Add branch predictor, BTB, load store queue, and storesets.
arch/isa_parser.py:
Specify the template parameter for AlphaDynInst
base/traceflags.py:
Add load store queue, store set, and mem dependence unit to the
list of trace flags.
cpu/base_dyn_inst.cc:
Change formating, add in debug statement.
cpu/base_dyn_inst.hh:
Change DynInst to be RefCounted, add flag to clear whether or not this
instruction can commit. This is likely to be removed in the future.
cpu/beta_cpu/alpha_dyn_inst.cc:
AlphaDynInst has been changed to be templated, so now this CC file
is just used to force instantiations of AlphaDynInst.
cpu/beta_cpu/alpha_dyn_inst.hh:
Changed AlphaDynInst to be templated on Impl. Removed some unnecessary
functions.
cpu/beta_cpu/alpha_full_cpu.cc:
AlphaFullCPU has been changed to be templated, so this CC file is now
just used to force instantation of AlphaFullCPU.
cpu/beta_cpu/alpha_full_cpu.hh:
Change AlphaFullCPU to be templated on Impl.
cpu/beta_cpu/alpha_impl.hh:
Update it to reflect AlphaDynInst and AlphaFullCPU being templated
on Impl. Also removed time buffers from here, as they are really
a part of the CPU and are thus in the CPU policy now.
cpu/beta_cpu/alpha_params.hh:
Make AlphaSimpleParams inherit from the BaseFullCPU so that it doesn't
need to specifically declare any parameters that are already in the
BaseFullCPU.
cpu/beta_cpu/comm.hh:
Changed the structure of the time buffer communication structs. Now
they include the size of the packet of instructions it is sending.
Added some parameters to the backwards communication struct, mainly
for squashing.
cpu/beta_cpu/commit.hh:
Update typenames to reflect change in location of time buffer structs.
Update DynInst to DynInstPtr (it is refcounted now).
cpu/beta_cpu/commit_impl.hh:
Formatting changes mainly. Also sends back proper information
on branch mispredicts so that the bpred unit can update itself.
Updated behavior for non-speculative instructions (stores, any
other non-spec instructions): once they reach the head of the ROB,
the ROB signals back to the IQ that it can go ahead and issue the
non-speculative instruction. The instruction itself is updated so that
commit won't try to commit it again until it is done executing.
cpu/beta_cpu/cpu_policy.hh:
Added branch prediction unit, mem dependence prediction unit, load
store queue. Moved time buffer structs from AlphaSimpleImpl to here.
cpu/beta_cpu/decode.hh:
Changed typedefs to reflect change in location of time buffer structs
and also the change from DynInst to ref counted DynInstPtr.
cpu/beta_cpu/decode_impl.hh:
Continues to buffer instructions even while unblocking now. Changed
how it loops through groups of instructions so it can properly block
during the middle of a group of instructions.
cpu/beta_cpu/fetch.hh:
Changed typedefs to reflect change in location of time buffer structs
and the change to ref counted DynInsts. Also added in branch
brediction unit.
cpu/beta_cpu/fetch_impl.hh:
Add in branch prediction. Changed how fetch checks inputs and its
current state to make for easier logic.
cpu/beta_cpu/free_list.cc:
Changed int regs and float regs to logically use one flat namespace.
Future change will be moving them to a single scoreboard to conserve
space.
cpu/beta_cpu/free_list.hh:
Mostly debugging statements. Might be removed for performance in future.
cpu/beta_cpu/full_cpu.cc:
Added in some debugging statements. Updated BaseFullCPU to take
a params object.
cpu/beta_cpu/full_cpu.hh:
Added params class within BaseCPU that other param classes will be
able to inherit from. Updated typedefs to reflect change in location
of time buffer structs and ref counted DynInst.
cpu/beta_cpu/iew.hh:
Updated typedefs to reflect change in location of time buffer structs
and use of ref counted DynInsts.
cpu/beta_cpu/iew_impl.hh:
Added in load store queue, updated iew to be able to execute non-
speculative instructions, instead of having them execute in commit.
cpu/beta_cpu/inst_queue.hh:
Updated change to ref counted DynInsts. Changed inst queue to hold
non-speculative instructions as well, which are issued only when
commit signals backwards that a nonspeculative instruction is at
the head of the ROB.
cpu/beta_cpu/inst_queue_impl.hh:
Updated to allow for non-speculative instructions to be in the inst
queue. Also added some debug functions.
cpu/beta_cpu/regfile.hh:
Added debugging statements, changed formatting.
cpu/beta_cpu/rename.hh:
Updated typedefs, added some functions to clean up code.
cpu/beta_cpu/rename_impl.hh:
Moved some code into functions to make it easier to read.
cpu/beta_cpu/rename_map.cc:
Changed int and float reg behavior to use a single flat namespace. In
the future, the rename maps can be combined to a single rename map to
save space.
cpu/beta_cpu/rename_map.hh:
Added destructor.
cpu/beta_cpu/rob.hh:
Updated it with change from DynInst to ref counted DynInst.
cpu/beta_cpu/rob_impl.hh:
Formatting, updated to use ref counted DynInst.
cpu/static_inst.hh:
Updated forward declaration for AlphaDynInst now that it is templated.
--HG--
extra : convert_revision : 1045f240ee9b6a4bd368e1806aca029ebbdc6dd3
2004-09-23 20:06:03 +02:00
|
|
|
|
|
|
|
// Consider this instruction executed so that commit can go
|
|
|
|
// ahead and retire the instruction.
|
|
|
|
inst->setExecuted();
|
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
// Not sure if I should set this here or just let commit try to
|
|
|
|
// commit any squashed instructions. I like the latter a bit more.
|
|
|
|
inst->setCanCommit();
|
Update to make multiple instruction issue and different latencies work.
Also change to ref counted DynInst.
SConscript:
Add branch predictor, BTB, load store queue, and storesets.
arch/isa_parser.py:
Specify the template parameter for AlphaDynInst
base/traceflags.py:
Add load store queue, store set, and mem dependence unit to the
list of trace flags.
cpu/base_dyn_inst.cc:
Change formating, add in debug statement.
cpu/base_dyn_inst.hh:
Change DynInst to be RefCounted, add flag to clear whether or not this
instruction can commit. This is likely to be removed in the future.
cpu/beta_cpu/alpha_dyn_inst.cc:
AlphaDynInst has been changed to be templated, so now this CC file
is just used to force instantiations of AlphaDynInst.
cpu/beta_cpu/alpha_dyn_inst.hh:
Changed AlphaDynInst to be templated on Impl. Removed some unnecessary
functions.
cpu/beta_cpu/alpha_full_cpu.cc:
AlphaFullCPU has been changed to be templated, so this CC file is now
just used to force instantation of AlphaFullCPU.
cpu/beta_cpu/alpha_full_cpu.hh:
Change AlphaFullCPU to be templated on Impl.
cpu/beta_cpu/alpha_impl.hh:
Update it to reflect AlphaDynInst and AlphaFullCPU being templated
on Impl. Also removed time buffers from here, as they are really
a part of the CPU and are thus in the CPU policy now.
cpu/beta_cpu/alpha_params.hh:
Make AlphaSimpleParams inherit from the BaseFullCPU so that it doesn't
need to specifically declare any parameters that are already in the
BaseFullCPU.
cpu/beta_cpu/comm.hh:
Changed the structure of the time buffer communication structs. Now
they include the size of the packet of instructions it is sending.
Added some parameters to the backwards communication struct, mainly
for squashing.
cpu/beta_cpu/commit.hh:
Update typenames to reflect change in location of time buffer structs.
Update DynInst to DynInstPtr (it is refcounted now).
cpu/beta_cpu/commit_impl.hh:
Formatting changes mainly. Also sends back proper information
on branch mispredicts so that the bpred unit can update itself.
Updated behavior for non-speculative instructions (stores, any
other non-spec instructions): once they reach the head of the ROB,
the ROB signals back to the IQ that it can go ahead and issue the
non-speculative instruction. The instruction itself is updated so that
commit won't try to commit it again until it is done executing.
cpu/beta_cpu/cpu_policy.hh:
Added branch prediction unit, mem dependence prediction unit, load
store queue. Moved time buffer structs from AlphaSimpleImpl to here.
cpu/beta_cpu/decode.hh:
Changed typedefs to reflect change in location of time buffer structs
and also the change from DynInst to ref counted DynInstPtr.
cpu/beta_cpu/decode_impl.hh:
Continues to buffer instructions even while unblocking now. Changed
how it loops through groups of instructions so it can properly block
during the middle of a group of instructions.
cpu/beta_cpu/fetch.hh:
Changed typedefs to reflect change in location of time buffer structs
and the change to ref counted DynInsts. Also added in branch
brediction unit.
cpu/beta_cpu/fetch_impl.hh:
Add in branch prediction. Changed how fetch checks inputs and its
current state to make for easier logic.
cpu/beta_cpu/free_list.cc:
Changed int regs and float regs to logically use one flat namespace.
Future change will be moving them to a single scoreboard to conserve
space.
cpu/beta_cpu/free_list.hh:
Mostly debugging statements. Might be removed for performance in future.
cpu/beta_cpu/full_cpu.cc:
Added in some debugging statements. Updated BaseFullCPU to take
a params object.
cpu/beta_cpu/full_cpu.hh:
Added params class within BaseCPU that other param classes will be
able to inherit from. Updated typedefs to reflect change in location
of time buffer structs and ref counted DynInst.
cpu/beta_cpu/iew.hh:
Updated typedefs to reflect change in location of time buffer structs
and use of ref counted DynInsts.
cpu/beta_cpu/iew_impl.hh:
Added in load store queue, updated iew to be able to execute non-
speculative instructions, instead of having them execute in commit.
cpu/beta_cpu/inst_queue.hh:
Updated change to ref counted DynInsts. Changed inst queue to hold
non-speculative instructions as well, which are issued only when
commit signals backwards that a nonspeculative instruction is at
the head of the ROB.
cpu/beta_cpu/inst_queue_impl.hh:
Updated to allow for non-speculative instructions to be in the inst
queue. Also added some debug functions.
cpu/beta_cpu/regfile.hh:
Added debugging statements, changed formatting.
cpu/beta_cpu/rename.hh:
Updated typedefs, added some functions to clean up code.
cpu/beta_cpu/rename_impl.hh:
Moved some code into functions to make it easier to read.
cpu/beta_cpu/rename_map.cc:
Changed int and float reg behavior to use a single flat namespace. In
the future, the rename maps can be combined to a single rename map to
save space.
cpu/beta_cpu/rename_map.hh:
Added destructor.
cpu/beta_cpu/rob.hh:
Updated it with change from DynInst to ref counted DynInst.
cpu/beta_cpu/rob_impl.hh:
Formatting, updated to use ref counted DynInst.
cpu/static_inst.hh:
Updated forward declaration for AlphaDynInst now that it is templated.
--HG--
extra : convert_revision : 1045f240ee9b6a4bd368e1806aca029ebbdc6dd3
2004-09-23 20:06:03 +02:00
|
|
|
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
++iewExecSquashedInsts;
|
|
|
|
|
2006-07-05 21:51:36 +02:00
|
|
|
decrWb(inst->seqNum);
|
2004-08-20 20:54:07 +02:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
Fault fault = NoFault;
|
2004-08-20 20:54:07 +02:00
|
|
|
|
|
|
|
// Execute instruction.
|
|
|
|
// Note that if the instruction faults, it will be handled
|
|
|
|
// at the commit stage.
|
2011-01-18 23:30:02 +01:00
|
|
|
if (inst->isMemRef()) {
|
2006-04-23 00:26:48 +02:00
|
|
|
DPRINTF(IEW, "Execute: Calculating address for memory "
|
Update to make multiple instruction issue and different latencies work.
Also change to ref counted DynInst.
SConscript:
Add branch predictor, BTB, load store queue, and storesets.
arch/isa_parser.py:
Specify the template parameter for AlphaDynInst
base/traceflags.py:
Add load store queue, store set, and mem dependence unit to the
list of trace flags.
cpu/base_dyn_inst.cc:
Change formating, add in debug statement.
cpu/base_dyn_inst.hh:
Change DynInst to be RefCounted, add flag to clear whether or not this
instruction can commit. This is likely to be removed in the future.
cpu/beta_cpu/alpha_dyn_inst.cc:
AlphaDynInst has been changed to be templated, so now this CC file
is just used to force instantiations of AlphaDynInst.
cpu/beta_cpu/alpha_dyn_inst.hh:
Changed AlphaDynInst to be templated on Impl. Removed some unnecessary
functions.
cpu/beta_cpu/alpha_full_cpu.cc:
AlphaFullCPU has been changed to be templated, so this CC file is now
just used to force instantation of AlphaFullCPU.
cpu/beta_cpu/alpha_full_cpu.hh:
Change AlphaFullCPU to be templated on Impl.
cpu/beta_cpu/alpha_impl.hh:
Update it to reflect AlphaDynInst and AlphaFullCPU being templated
on Impl. Also removed time buffers from here, as they are really
a part of the CPU and are thus in the CPU policy now.
cpu/beta_cpu/alpha_params.hh:
Make AlphaSimpleParams inherit from the BaseFullCPU so that it doesn't
need to specifically declare any parameters that are already in the
BaseFullCPU.
cpu/beta_cpu/comm.hh:
Changed the structure of the time buffer communication structs. Now
they include the size of the packet of instructions it is sending.
Added some parameters to the backwards communication struct, mainly
for squashing.
cpu/beta_cpu/commit.hh:
Update typenames to reflect change in location of time buffer structs.
Update DynInst to DynInstPtr (it is refcounted now).
cpu/beta_cpu/commit_impl.hh:
Formatting changes mainly. Also sends back proper information
on branch mispredicts so that the bpred unit can update itself.
Updated behavior for non-speculative instructions (stores, any
other non-spec instructions): once they reach the head of the ROB,
the ROB signals back to the IQ that it can go ahead and issue the
non-speculative instruction. The instruction itself is updated so that
commit won't try to commit it again until it is done executing.
cpu/beta_cpu/cpu_policy.hh:
Added branch prediction unit, mem dependence prediction unit, load
store queue. Moved time buffer structs from AlphaSimpleImpl to here.
cpu/beta_cpu/decode.hh:
Changed typedefs to reflect change in location of time buffer structs
and also the change from DynInst to ref counted DynInstPtr.
cpu/beta_cpu/decode_impl.hh:
Continues to buffer instructions even while unblocking now. Changed
how it loops through groups of instructions so it can properly block
during the middle of a group of instructions.
cpu/beta_cpu/fetch.hh:
Changed typedefs to reflect change in location of time buffer structs
and the change to ref counted DynInsts. Also added in branch
brediction unit.
cpu/beta_cpu/fetch_impl.hh:
Add in branch prediction. Changed how fetch checks inputs and its
current state to make for easier logic.
cpu/beta_cpu/free_list.cc:
Changed int regs and float regs to logically use one flat namespace.
Future change will be moving them to a single scoreboard to conserve
space.
cpu/beta_cpu/free_list.hh:
Mostly debugging statements. Might be removed for performance in future.
cpu/beta_cpu/full_cpu.cc:
Added in some debugging statements. Updated BaseFullCPU to take
a params object.
cpu/beta_cpu/full_cpu.hh:
Added params class within BaseCPU that other param classes will be
able to inherit from. Updated typedefs to reflect change in location
of time buffer structs and ref counted DynInst.
cpu/beta_cpu/iew.hh:
Updated typedefs to reflect change in location of time buffer structs
and use of ref counted DynInsts.
cpu/beta_cpu/iew_impl.hh:
Added in load store queue, updated iew to be able to execute non-
speculative instructions, instead of having them execute in commit.
cpu/beta_cpu/inst_queue.hh:
Updated change to ref counted DynInsts. Changed inst queue to hold
non-speculative instructions as well, which are issued only when
commit signals backwards that a nonspeculative instruction is at
the head of the ROB.
cpu/beta_cpu/inst_queue_impl.hh:
Updated to allow for non-speculative instructions to be in the inst
queue. Also added some debug functions.
cpu/beta_cpu/regfile.hh:
Added debugging statements, changed formatting.
cpu/beta_cpu/rename.hh:
Updated typedefs, added some functions to clean up code.
cpu/beta_cpu/rename_impl.hh:
Moved some code into functions to make it easier to read.
cpu/beta_cpu/rename_map.cc:
Changed int and float reg behavior to use a single flat namespace. In
the future, the rename maps can be combined to a single rename map to
save space.
cpu/beta_cpu/rename_map.hh:
Added destructor.
cpu/beta_cpu/rob.hh:
Updated it with change from DynInst to ref counted DynInst.
cpu/beta_cpu/rob_impl.hh:
Formatting, updated to use ref counted DynInst.
cpu/static_inst.hh:
Updated forward declaration for AlphaDynInst now that it is templated.
--HG--
extra : convert_revision : 1045f240ee9b6a4bd368e1806aca029ebbdc6dd3
2004-09-23 20:06:03 +02:00
|
|
|
"reference.\n");
|
|
|
|
|
|
|
|
// Tell the LDSTQ to execute this instruction (if it is a load).
|
|
|
|
if (inst->isLoad()) {
|
2006-04-23 00:26:48 +02:00
|
|
|
// Loads will mark themselves as executed, and their writeback
|
|
|
|
// event adds the instruction to the queue to commit
|
|
|
|
fault = ldstQueue.executeLoad(inst);
|
2011-02-12 01:29:35 +01:00
|
|
|
|
|
|
|
if (inst->isTranslationDelayed() &&
|
|
|
|
fault == NoFault) {
|
|
|
|
// A hw page table walk is currently going on; the
|
|
|
|
// instruction must be deferred.
|
|
|
|
DPRINTF(IEW, "Execute: Delayed translation, deferring "
|
|
|
|
"load.\n");
|
|
|
|
instQueue.deferMemInst(inst);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2011-01-18 23:30:02 +01:00
|
|
|
if (inst->isDataPrefetch() || inst->isInstPrefetch()) {
|
2011-02-23 22:10:50 +01:00
|
|
|
inst->fault = NoFault;
|
2011-01-18 23:30:02 +01:00
|
|
|
}
|
Update to make multiple instruction issue and different latencies work.
Also change to ref counted DynInst.
SConscript:
Add branch predictor, BTB, load store queue, and storesets.
arch/isa_parser.py:
Specify the template parameter for AlphaDynInst
base/traceflags.py:
Add load store queue, store set, and mem dependence unit to the
list of trace flags.
cpu/base_dyn_inst.cc:
Change formating, add in debug statement.
cpu/base_dyn_inst.hh:
Change DynInst to be RefCounted, add flag to clear whether or not this
instruction can commit. This is likely to be removed in the future.
cpu/beta_cpu/alpha_dyn_inst.cc:
AlphaDynInst has been changed to be templated, so now this CC file
is just used to force instantiations of AlphaDynInst.
cpu/beta_cpu/alpha_dyn_inst.hh:
Changed AlphaDynInst to be templated on Impl. Removed some unnecessary
functions.
cpu/beta_cpu/alpha_full_cpu.cc:
AlphaFullCPU has been changed to be templated, so this CC file is now
just used to force instantation of AlphaFullCPU.
cpu/beta_cpu/alpha_full_cpu.hh:
Change AlphaFullCPU to be templated on Impl.
cpu/beta_cpu/alpha_impl.hh:
Update it to reflect AlphaDynInst and AlphaFullCPU being templated
on Impl. Also removed time buffers from here, as they are really
a part of the CPU and are thus in the CPU policy now.
cpu/beta_cpu/alpha_params.hh:
Make AlphaSimpleParams inherit from the BaseFullCPU so that it doesn't
need to specifically declare any parameters that are already in the
BaseFullCPU.
cpu/beta_cpu/comm.hh:
Changed the structure of the time buffer communication structs. Now
they include the size of the packet of instructions it is sending.
Added some parameters to the backwards communication struct, mainly
for squashing.
cpu/beta_cpu/commit.hh:
Update typenames to reflect change in location of time buffer structs.
Update DynInst to DynInstPtr (it is refcounted now).
cpu/beta_cpu/commit_impl.hh:
Formatting changes mainly. Also sends back proper information
on branch mispredicts so that the bpred unit can update itself.
Updated behavior for non-speculative instructions (stores, any
other non-spec instructions): once they reach the head of the ROB,
the ROB signals back to the IQ that it can go ahead and issue the
non-speculative instruction. The instruction itself is updated so that
commit won't try to commit it again until it is done executing.
cpu/beta_cpu/cpu_policy.hh:
Added branch prediction unit, mem dependence prediction unit, load
store queue. Moved time buffer structs from AlphaSimpleImpl to here.
cpu/beta_cpu/decode.hh:
Changed typedefs to reflect change in location of time buffer structs
and also the change from DynInst to ref counted DynInstPtr.
cpu/beta_cpu/decode_impl.hh:
Continues to buffer instructions even while unblocking now. Changed
how it loops through groups of instructions so it can properly block
during the middle of a group of instructions.
cpu/beta_cpu/fetch.hh:
Changed typedefs to reflect change in location of time buffer structs
and the change to ref counted DynInsts. Also added in branch
brediction unit.
cpu/beta_cpu/fetch_impl.hh:
Add in branch prediction. Changed how fetch checks inputs and its
current state to make for easier logic.
cpu/beta_cpu/free_list.cc:
Changed int regs and float regs to logically use one flat namespace.
Future change will be moving them to a single scoreboard to conserve
space.
cpu/beta_cpu/free_list.hh:
Mostly debugging statements. Might be removed for performance in future.
cpu/beta_cpu/full_cpu.cc:
Added in some debugging statements. Updated BaseFullCPU to take
a params object.
cpu/beta_cpu/full_cpu.hh:
Added params class within BaseCPU that other param classes will be
able to inherit from. Updated typedefs to reflect change in location
of time buffer structs and ref counted DynInst.
cpu/beta_cpu/iew.hh:
Updated typedefs to reflect change in location of time buffer structs
and use of ref counted DynInsts.
cpu/beta_cpu/iew_impl.hh:
Added in load store queue, updated iew to be able to execute non-
speculative instructions, instead of having them execute in commit.
cpu/beta_cpu/inst_queue.hh:
Updated change to ref counted DynInsts. Changed inst queue to hold
non-speculative instructions as well, which are issued only when
commit signals backwards that a nonspeculative instruction is at
the head of the ROB.
cpu/beta_cpu/inst_queue_impl.hh:
Updated to allow for non-speculative instructions to be in the inst
queue. Also added some debug functions.
cpu/beta_cpu/regfile.hh:
Added debugging statements, changed formatting.
cpu/beta_cpu/rename.hh:
Updated typedefs, added some functions to clean up code.
cpu/beta_cpu/rename_impl.hh:
Moved some code into functions to make it easier to read.
cpu/beta_cpu/rename_map.cc:
Changed int and float reg behavior to use a single flat namespace. In
the future, the rename maps can be combined to a single rename map to
save space.
cpu/beta_cpu/rename_map.hh:
Added destructor.
cpu/beta_cpu/rob.hh:
Updated it with change from DynInst to ref counted DynInst.
cpu/beta_cpu/rob_impl.hh:
Formatting, updated to use ref counted DynInst.
cpu/static_inst.hh:
Updated forward declaration for AlphaDynInst now that it is templated.
--HG--
extra : convert_revision : 1045f240ee9b6a4bd368e1806aca029ebbdc6dd3
2004-09-23 20:06:03 +02:00
|
|
|
} else if (inst->isStore()) {
|
2006-09-28 06:09:27 +02:00
|
|
|
fault = ldstQueue.executeStore(inst);
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2011-02-12 01:29:35 +01:00
|
|
|
if (inst->isTranslationDelayed() &&
|
|
|
|
fault == NoFault) {
|
|
|
|
// A hw page table walk is currently going on; the
|
|
|
|
// instruction must be deferred.
|
|
|
|
DPRINTF(IEW, "Execute: Delayed translation, deferring "
|
|
|
|
"store.\n");
|
|
|
|
instQueue.deferMemInst(inst);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
// If the store had a fault then it may not have a mem req
|
2010-12-08 01:19:57 +01:00
|
|
|
if (fault != NoFault || inst->readPredicate() == false ||
|
|
|
|
!inst->isStoreConditional()) {
|
|
|
|
// If the instruction faulted, then we need to send it along
|
|
|
|
// to commit without the instruction completing.
|
2006-09-28 06:09:27 +02:00
|
|
|
// Send this instruction to commit, also make sure iew stage
|
|
|
|
// realizes there is activity.
|
2006-04-23 00:26:48 +02:00
|
|
|
inst->setExecuted();
|
|
|
|
instToCommit(inst);
|
2006-09-28 06:09:27 +02:00
|
|
|
activityThisCycle();
|
2006-04-23 00:26:48 +02:00
|
|
|
}
|
2006-05-19 21:44:03 +02:00
|
|
|
|
|
|
|
// Store conditionals will mark themselves as
|
|
|
|
// executed, and their writeback event will add the
|
|
|
|
// instruction to the queue to commit.
|
Update to make multiple instruction issue and different latencies work.
Also change to ref counted DynInst.
SConscript:
Add branch predictor, BTB, load store queue, and storesets.
arch/isa_parser.py:
Specify the template parameter for AlphaDynInst
base/traceflags.py:
Add load store queue, store set, and mem dependence unit to the
list of trace flags.
cpu/base_dyn_inst.cc:
Change formating, add in debug statement.
cpu/base_dyn_inst.hh:
Change DynInst to be RefCounted, add flag to clear whether or not this
instruction can commit. This is likely to be removed in the future.
cpu/beta_cpu/alpha_dyn_inst.cc:
AlphaDynInst has been changed to be templated, so now this CC file
is just used to force instantiations of AlphaDynInst.
cpu/beta_cpu/alpha_dyn_inst.hh:
Changed AlphaDynInst to be templated on Impl. Removed some unnecessary
functions.
cpu/beta_cpu/alpha_full_cpu.cc:
AlphaFullCPU has been changed to be templated, so this CC file is now
just used to force instantation of AlphaFullCPU.
cpu/beta_cpu/alpha_full_cpu.hh:
Change AlphaFullCPU to be templated on Impl.
cpu/beta_cpu/alpha_impl.hh:
Update it to reflect AlphaDynInst and AlphaFullCPU being templated
on Impl. Also removed time buffers from here, as they are really
a part of the CPU and are thus in the CPU policy now.
cpu/beta_cpu/alpha_params.hh:
Make AlphaSimpleParams inherit from the BaseFullCPU so that it doesn't
need to specifically declare any parameters that are already in the
BaseFullCPU.
cpu/beta_cpu/comm.hh:
Changed the structure of the time buffer communication structs. Now
they include the size of the packet of instructions it is sending.
Added some parameters to the backwards communication struct, mainly
for squashing.
cpu/beta_cpu/commit.hh:
Update typenames to reflect change in location of time buffer structs.
Update DynInst to DynInstPtr (it is refcounted now).
cpu/beta_cpu/commit_impl.hh:
Formatting changes mainly. Also sends back proper information
on branch mispredicts so that the bpred unit can update itself.
Updated behavior for non-speculative instructions (stores, any
other non-spec instructions): once they reach the head of the ROB,
the ROB signals back to the IQ that it can go ahead and issue the
non-speculative instruction. The instruction itself is updated so that
commit won't try to commit it again until it is done executing.
cpu/beta_cpu/cpu_policy.hh:
Added branch prediction unit, mem dependence prediction unit, load
store queue. Moved time buffer structs from AlphaSimpleImpl to here.
cpu/beta_cpu/decode.hh:
Changed typedefs to reflect change in location of time buffer structs
and also the change from DynInst to ref counted DynInstPtr.
cpu/beta_cpu/decode_impl.hh:
Continues to buffer instructions even while unblocking now. Changed
how it loops through groups of instructions so it can properly block
during the middle of a group of instructions.
cpu/beta_cpu/fetch.hh:
Changed typedefs to reflect change in location of time buffer structs
and the change to ref counted DynInsts. Also added in branch
brediction unit.
cpu/beta_cpu/fetch_impl.hh:
Add in branch prediction. Changed how fetch checks inputs and its
current state to make for easier logic.
cpu/beta_cpu/free_list.cc:
Changed int regs and float regs to logically use one flat namespace.
Future change will be moving them to a single scoreboard to conserve
space.
cpu/beta_cpu/free_list.hh:
Mostly debugging statements. Might be removed for performance in future.
cpu/beta_cpu/full_cpu.cc:
Added in some debugging statements. Updated BaseFullCPU to take
a params object.
cpu/beta_cpu/full_cpu.hh:
Added params class within BaseCPU that other param classes will be
able to inherit from. Updated typedefs to reflect change in location
of time buffer structs and ref counted DynInst.
cpu/beta_cpu/iew.hh:
Updated typedefs to reflect change in location of time buffer structs
and use of ref counted DynInsts.
cpu/beta_cpu/iew_impl.hh:
Added in load store queue, updated iew to be able to execute non-
speculative instructions, instead of having them execute in commit.
cpu/beta_cpu/inst_queue.hh:
Updated change to ref counted DynInsts. Changed inst queue to hold
non-speculative instructions as well, which are issued only when
commit signals backwards that a nonspeculative instruction is at
the head of the ROB.
cpu/beta_cpu/inst_queue_impl.hh:
Updated to allow for non-speculative instructions to be in the inst
queue. Also added some debug functions.
cpu/beta_cpu/regfile.hh:
Added debugging statements, changed formatting.
cpu/beta_cpu/rename.hh:
Updated typedefs, added some functions to clean up code.
cpu/beta_cpu/rename_impl.hh:
Moved some code into functions to make it easier to read.
cpu/beta_cpu/rename_map.cc:
Changed int and float reg behavior to use a single flat namespace. In
the future, the rename maps can be combined to a single rename map to
save space.
cpu/beta_cpu/rename_map.hh:
Added destructor.
cpu/beta_cpu/rob.hh:
Updated it with change from DynInst to ref counted DynInst.
cpu/beta_cpu/rob_impl.hh:
Formatting, updated to use ref counted DynInst.
cpu/static_inst.hh:
Updated forward declaration for AlphaDynInst now that it is templated.
--HG--
extra : convert_revision : 1045f240ee9b6a4bd368e1806aca029ebbdc6dd3
2004-09-23 20:06:03 +02:00
|
|
|
} else {
|
2006-04-23 00:26:48 +02:00
|
|
|
panic("Unexpected memory type!\n");
|
Update to make multiple instruction issue and different latencies work.
Also change to ref counted DynInst.
SConscript:
Add branch predictor, BTB, load store queue, and storesets.
arch/isa_parser.py:
Specify the template parameter for AlphaDynInst
base/traceflags.py:
Add load store queue, store set, and mem dependence unit to the
list of trace flags.
cpu/base_dyn_inst.cc:
Change formating, add in debug statement.
cpu/base_dyn_inst.hh:
Change DynInst to be RefCounted, add flag to clear whether or not this
instruction can commit. This is likely to be removed in the future.
cpu/beta_cpu/alpha_dyn_inst.cc:
AlphaDynInst has been changed to be templated, so now this CC file
is just used to force instantiations of AlphaDynInst.
cpu/beta_cpu/alpha_dyn_inst.hh:
Changed AlphaDynInst to be templated on Impl. Removed some unnecessary
functions.
cpu/beta_cpu/alpha_full_cpu.cc:
AlphaFullCPU has been changed to be templated, so this CC file is now
just used to force instantation of AlphaFullCPU.
cpu/beta_cpu/alpha_full_cpu.hh:
Change AlphaFullCPU to be templated on Impl.
cpu/beta_cpu/alpha_impl.hh:
Update it to reflect AlphaDynInst and AlphaFullCPU being templated
on Impl. Also removed time buffers from here, as they are really
a part of the CPU and are thus in the CPU policy now.
cpu/beta_cpu/alpha_params.hh:
Make AlphaSimpleParams inherit from the BaseFullCPU so that it doesn't
need to specifically declare any parameters that are already in the
BaseFullCPU.
cpu/beta_cpu/comm.hh:
Changed the structure of the time buffer communication structs. Now
they include the size of the packet of instructions it is sending.
Added some parameters to the backwards communication struct, mainly
for squashing.
cpu/beta_cpu/commit.hh:
Update typenames to reflect change in location of time buffer structs.
Update DynInst to DynInstPtr (it is refcounted now).
cpu/beta_cpu/commit_impl.hh:
Formatting changes mainly. Also sends back proper information
on branch mispredicts so that the bpred unit can update itself.
Updated behavior for non-speculative instructions (stores, any
other non-spec instructions): once they reach the head of the ROB,
the ROB signals back to the IQ that it can go ahead and issue the
non-speculative instruction. The instruction itself is updated so that
commit won't try to commit it again until it is done executing.
cpu/beta_cpu/cpu_policy.hh:
Added branch prediction unit, mem dependence prediction unit, load
store queue. Moved time buffer structs from AlphaSimpleImpl to here.
cpu/beta_cpu/decode.hh:
Changed typedefs to reflect change in location of time buffer structs
and also the change from DynInst to ref counted DynInstPtr.
cpu/beta_cpu/decode_impl.hh:
Continues to buffer instructions even while unblocking now. Changed
how it loops through groups of instructions so it can properly block
during the middle of a group of instructions.
cpu/beta_cpu/fetch.hh:
Changed typedefs to reflect change in location of time buffer structs
and the change to ref counted DynInsts. Also added in branch
brediction unit.
cpu/beta_cpu/fetch_impl.hh:
Add in branch prediction. Changed how fetch checks inputs and its
current state to make for easier logic.
cpu/beta_cpu/free_list.cc:
Changed int regs and float regs to logically use one flat namespace.
Future change will be moving them to a single scoreboard to conserve
space.
cpu/beta_cpu/free_list.hh:
Mostly debugging statements. Might be removed for performance in future.
cpu/beta_cpu/full_cpu.cc:
Added in some debugging statements. Updated BaseFullCPU to take
a params object.
cpu/beta_cpu/full_cpu.hh:
Added params class within BaseCPU that other param classes will be
able to inherit from. Updated typedefs to reflect change in location
of time buffer structs and ref counted DynInst.
cpu/beta_cpu/iew.hh:
Updated typedefs to reflect change in location of time buffer structs
and use of ref counted DynInsts.
cpu/beta_cpu/iew_impl.hh:
Added in load store queue, updated iew to be able to execute non-
speculative instructions, instead of having them execute in commit.
cpu/beta_cpu/inst_queue.hh:
Updated change to ref counted DynInsts. Changed inst queue to hold
non-speculative instructions as well, which are issued only when
commit signals backwards that a nonspeculative instruction is at
the head of the ROB.
cpu/beta_cpu/inst_queue_impl.hh:
Updated to allow for non-speculative instructions to be in the inst
queue. Also added some debug functions.
cpu/beta_cpu/regfile.hh:
Added debugging statements, changed formatting.
cpu/beta_cpu/rename.hh:
Updated typedefs, added some functions to clean up code.
cpu/beta_cpu/rename_impl.hh:
Moved some code into functions to make it easier to read.
cpu/beta_cpu/rename_map.cc:
Changed int and float reg behavior to use a single flat namespace. In
the future, the rename maps can be combined to a single rename map to
save space.
cpu/beta_cpu/rename_map.hh:
Added destructor.
cpu/beta_cpu/rob.hh:
Updated it with change from DynInst to ref counted DynInst.
cpu/beta_cpu/rob_impl.hh:
Formatting, updated to use ref counted DynInst.
cpu/static_inst.hh:
Updated forward declaration for AlphaDynInst now that it is templated.
--HG--
extra : convert_revision : 1045f240ee9b6a4bd368e1806aca029ebbdc6dd3
2004-09-23 20:06:03 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
} else {
|
2011-01-18 23:30:01 +01:00
|
|
|
// If the instruction has already faulted, then skip executing it.
|
|
|
|
// Such case can happen when it faulted during ITLB translation.
|
|
|
|
// If we execute the instruction (even if it's a nop) the fault
|
|
|
|
// will be replaced and we will lose it.
|
|
|
|
if (inst->getFault() == NoFault) {
|
|
|
|
inst->execute();
|
2011-01-18 23:30:02 +01:00
|
|
|
if (inst->readPredicate() == false)
|
|
|
|
inst->forwardOldRegs();
|
2011-01-18 23:30:01 +01:00
|
|
|
}
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
inst->setExecuted();
|
2005-05-03 16:56:47 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
instToCommit(inst);
|
2005-05-03 16:56:47 +02:00
|
|
|
}
|
|
|
|
|
2006-04-24 23:06:00 +02:00
|
|
|
updateExeInstStats(inst);
|
|
|
|
|
2006-05-19 21:44:03 +02:00
|
|
|
// Check if branch prediction was correct, if not then we need
|
|
|
|
// to tell commit to squash in flight instructions. Only
|
|
|
|
// handle this if there hasn't already been something that
|
Update to make multiple instruction issue and different latencies work.
Also change to ref counted DynInst.
SConscript:
Add branch predictor, BTB, load store queue, and storesets.
arch/isa_parser.py:
Specify the template parameter for AlphaDynInst
base/traceflags.py:
Add load store queue, store set, and mem dependence unit to the
list of trace flags.
cpu/base_dyn_inst.cc:
Change formating, add in debug statement.
cpu/base_dyn_inst.hh:
Change DynInst to be RefCounted, add flag to clear whether or not this
instruction can commit. This is likely to be removed in the future.
cpu/beta_cpu/alpha_dyn_inst.cc:
AlphaDynInst has been changed to be templated, so now this CC file
is just used to force instantiations of AlphaDynInst.
cpu/beta_cpu/alpha_dyn_inst.hh:
Changed AlphaDynInst to be templated on Impl. Removed some unnecessary
functions.
cpu/beta_cpu/alpha_full_cpu.cc:
AlphaFullCPU has been changed to be templated, so this CC file is now
just used to force instantation of AlphaFullCPU.
cpu/beta_cpu/alpha_full_cpu.hh:
Change AlphaFullCPU to be templated on Impl.
cpu/beta_cpu/alpha_impl.hh:
Update it to reflect AlphaDynInst and AlphaFullCPU being templated
on Impl. Also removed time buffers from here, as they are really
a part of the CPU and are thus in the CPU policy now.
cpu/beta_cpu/alpha_params.hh:
Make AlphaSimpleParams inherit from the BaseFullCPU so that it doesn't
need to specifically declare any parameters that are already in the
BaseFullCPU.
cpu/beta_cpu/comm.hh:
Changed the structure of the time buffer communication structs. Now
they include the size of the packet of instructions it is sending.
Added some parameters to the backwards communication struct, mainly
for squashing.
cpu/beta_cpu/commit.hh:
Update typenames to reflect change in location of time buffer structs.
Update DynInst to DynInstPtr (it is refcounted now).
cpu/beta_cpu/commit_impl.hh:
Formatting changes mainly. Also sends back proper information
on branch mispredicts so that the bpred unit can update itself.
Updated behavior for non-speculative instructions (stores, any
other non-spec instructions): once they reach the head of the ROB,
the ROB signals back to the IQ that it can go ahead and issue the
non-speculative instruction. The instruction itself is updated so that
commit won't try to commit it again until it is done executing.
cpu/beta_cpu/cpu_policy.hh:
Added branch prediction unit, mem dependence prediction unit, load
store queue. Moved time buffer structs from AlphaSimpleImpl to here.
cpu/beta_cpu/decode.hh:
Changed typedefs to reflect change in location of time buffer structs
and also the change from DynInst to ref counted DynInstPtr.
cpu/beta_cpu/decode_impl.hh:
Continues to buffer instructions even while unblocking now. Changed
how it loops through groups of instructions so it can properly block
during the middle of a group of instructions.
cpu/beta_cpu/fetch.hh:
Changed typedefs to reflect change in location of time buffer structs
and the change to ref counted DynInsts. Also added in branch
brediction unit.
cpu/beta_cpu/fetch_impl.hh:
Add in branch prediction. Changed how fetch checks inputs and its
current state to make for easier logic.
cpu/beta_cpu/free_list.cc:
Changed int regs and float regs to logically use one flat namespace.
Future change will be moving them to a single scoreboard to conserve
space.
cpu/beta_cpu/free_list.hh:
Mostly debugging statements. Might be removed for performance in future.
cpu/beta_cpu/full_cpu.cc:
Added in some debugging statements. Updated BaseFullCPU to take
a params object.
cpu/beta_cpu/full_cpu.hh:
Added params class within BaseCPU that other param classes will be
able to inherit from. Updated typedefs to reflect change in location
of time buffer structs and ref counted DynInst.
cpu/beta_cpu/iew.hh:
Updated typedefs to reflect change in location of time buffer structs
and use of ref counted DynInsts.
cpu/beta_cpu/iew_impl.hh:
Added in load store queue, updated iew to be able to execute non-
speculative instructions, instead of having them execute in commit.
cpu/beta_cpu/inst_queue.hh:
Updated change to ref counted DynInsts. Changed inst queue to hold
non-speculative instructions as well, which are issued only when
commit signals backwards that a nonspeculative instruction is at
the head of the ROB.
cpu/beta_cpu/inst_queue_impl.hh:
Updated to allow for non-speculative instructions to be in the inst
queue. Also added some debug functions.
cpu/beta_cpu/regfile.hh:
Added debugging statements, changed formatting.
cpu/beta_cpu/rename.hh:
Updated typedefs, added some functions to clean up code.
cpu/beta_cpu/rename_impl.hh:
Moved some code into functions to make it easier to read.
cpu/beta_cpu/rename_map.cc:
Changed int and float reg behavior to use a single flat namespace. In
the future, the rename maps can be combined to a single rename map to
save space.
cpu/beta_cpu/rename_map.hh:
Added destructor.
cpu/beta_cpu/rob.hh:
Updated it with change from DynInst to ref counted DynInst.
cpu/beta_cpu/rob_impl.hh:
Formatting, updated to use ref counted DynInst.
cpu/static_inst.hh:
Updated forward declaration for AlphaDynInst now that it is templated.
--HG--
extra : convert_revision : 1045f240ee9b6a4bd368e1806aca029ebbdc6dd3
2004-09-23 20:06:03 +02:00
|
|
|
// redirects fetch in this group of instructions.
|
2006-04-23 00:26:48 +02:00
|
|
|
|
|
|
|
// 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.
|
2009-05-26 18:23:13 +02:00
|
|
|
ThreadID tid = inst->threadNumber;
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2006-12-12 23:35:46 +01:00
|
|
|
if (!fetchRedirect[tid] ||
|
2011-01-18 23:30:05 +01:00
|
|
|
!toCommit->squash[tid] ||
|
2006-12-12 23:35:46 +01:00
|
|
|
toCommit->squashedSeqNum[tid] > inst->seqNum) {
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2011-01-18 23:30:05 +01:00
|
|
|
// Prevent testing for misprediction on load instructions,
|
|
|
|
// that have not been executed.
|
|
|
|
bool loadNotExecuted = !inst->isExecuted() && inst->isLoad();
|
|
|
|
|
|
|
|
if (inst->mispredicted() && !loadNotExecuted) {
|
2006-04-23 00:26:48 +02:00
|
|
|
fetchRedirect[tid] = true;
|
Update to make multiple instruction issue and different latencies work.
Also change to ref counted DynInst.
SConscript:
Add branch predictor, BTB, load store queue, and storesets.
arch/isa_parser.py:
Specify the template parameter for AlphaDynInst
base/traceflags.py:
Add load store queue, store set, and mem dependence unit to the
list of trace flags.
cpu/base_dyn_inst.cc:
Change formating, add in debug statement.
cpu/base_dyn_inst.hh:
Change DynInst to be RefCounted, add flag to clear whether or not this
instruction can commit. This is likely to be removed in the future.
cpu/beta_cpu/alpha_dyn_inst.cc:
AlphaDynInst has been changed to be templated, so now this CC file
is just used to force instantiations of AlphaDynInst.
cpu/beta_cpu/alpha_dyn_inst.hh:
Changed AlphaDynInst to be templated on Impl. Removed some unnecessary
functions.
cpu/beta_cpu/alpha_full_cpu.cc:
AlphaFullCPU has been changed to be templated, so this CC file is now
just used to force instantation of AlphaFullCPU.
cpu/beta_cpu/alpha_full_cpu.hh:
Change AlphaFullCPU to be templated on Impl.
cpu/beta_cpu/alpha_impl.hh:
Update it to reflect AlphaDynInst and AlphaFullCPU being templated
on Impl. Also removed time buffers from here, as they are really
a part of the CPU and are thus in the CPU policy now.
cpu/beta_cpu/alpha_params.hh:
Make AlphaSimpleParams inherit from the BaseFullCPU so that it doesn't
need to specifically declare any parameters that are already in the
BaseFullCPU.
cpu/beta_cpu/comm.hh:
Changed the structure of the time buffer communication structs. Now
they include the size of the packet of instructions it is sending.
Added some parameters to the backwards communication struct, mainly
for squashing.
cpu/beta_cpu/commit.hh:
Update typenames to reflect change in location of time buffer structs.
Update DynInst to DynInstPtr (it is refcounted now).
cpu/beta_cpu/commit_impl.hh:
Formatting changes mainly. Also sends back proper information
on branch mispredicts so that the bpred unit can update itself.
Updated behavior for non-speculative instructions (stores, any
other non-spec instructions): once they reach the head of the ROB,
the ROB signals back to the IQ that it can go ahead and issue the
non-speculative instruction. The instruction itself is updated so that
commit won't try to commit it again until it is done executing.
cpu/beta_cpu/cpu_policy.hh:
Added branch prediction unit, mem dependence prediction unit, load
store queue. Moved time buffer structs from AlphaSimpleImpl to here.
cpu/beta_cpu/decode.hh:
Changed typedefs to reflect change in location of time buffer structs
and also the change from DynInst to ref counted DynInstPtr.
cpu/beta_cpu/decode_impl.hh:
Continues to buffer instructions even while unblocking now. Changed
how it loops through groups of instructions so it can properly block
during the middle of a group of instructions.
cpu/beta_cpu/fetch.hh:
Changed typedefs to reflect change in location of time buffer structs
and the change to ref counted DynInsts. Also added in branch
brediction unit.
cpu/beta_cpu/fetch_impl.hh:
Add in branch prediction. Changed how fetch checks inputs and its
current state to make for easier logic.
cpu/beta_cpu/free_list.cc:
Changed int regs and float regs to logically use one flat namespace.
Future change will be moving them to a single scoreboard to conserve
space.
cpu/beta_cpu/free_list.hh:
Mostly debugging statements. Might be removed for performance in future.
cpu/beta_cpu/full_cpu.cc:
Added in some debugging statements. Updated BaseFullCPU to take
a params object.
cpu/beta_cpu/full_cpu.hh:
Added params class within BaseCPU that other param classes will be
able to inherit from. Updated typedefs to reflect change in location
of time buffer structs and ref counted DynInst.
cpu/beta_cpu/iew.hh:
Updated typedefs to reflect change in location of time buffer structs
and use of ref counted DynInsts.
cpu/beta_cpu/iew_impl.hh:
Added in load store queue, updated iew to be able to execute non-
speculative instructions, instead of having them execute in commit.
cpu/beta_cpu/inst_queue.hh:
Updated change to ref counted DynInsts. Changed inst queue to hold
non-speculative instructions as well, which are issued only when
commit signals backwards that a nonspeculative instruction is at
the head of the ROB.
cpu/beta_cpu/inst_queue_impl.hh:
Updated to allow for non-speculative instructions to be in the inst
queue. Also added some debug functions.
cpu/beta_cpu/regfile.hh:
Added debugging statements, changed formatting.
cpu/beta_cpu/rename.hh:
Updated typedefs, added some functions to clean up code.
cpu/beta_cpu/rename_impl.hh:
Moved some code into functions to make it easier to read.
cpu/beta_cpu/rename_map.cc:
Changed int and float reg behavior to use a single flat namespace. In
the future, the rename maps can be combined to a single rename map to
save space.
cpu/beta_cpu/rename_map.hh:
Added destructor.
cpu/beta_cpu/rob.hh:
Updated it with change from DynInst to ref counted DynInst.
cpu/beta_cpu/rob_impl.hh:
Formatting, updated to use ref counted DynInst.
cpu/static_inst.hh:
Updated forward declaration for AlphaDynInst now that it is templated.
--HG--
extra : convert_revision : 1045f240ee9b6a4bd368e1806aca029ebbdc6dd3
2004-09-23 20:06:03 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
DPRINTF(IEW, "Execute: Branch mispredict detected.\n");
|
2009-04-18 16:42:29 +02:00
|
|
|
DPRINTF(IEW, "Predicted target was PC:%#x, NPC:%#x.\n",
|
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
|
|
|
inst->predInstAddr(), inst->predNextInstAddr());
|
|
|
|
DPRINTF(IEW, "Execute: Redirecting fetch to PC: %s.\n",
|
|
|
|
inst->pcState(), inst->nextInstAddr());
|
Update to make multiple instruction issue and different latencies work.
Also change to ref counted DynInst.
SConscript:
Add branch predictor, BTB, load store queue, and storesets.
arch/isa_parser.py:
Specify the template parameter for AlphaDynInst
base/traceflags.py:
Add load store queue, store set, and mem dependence unit to the
list of trace flags.
cpu/base_dyn_inst.cc:
Change formating, add in debug statement.
cpu/base_dyn_inst.hh:
Change DynInst to be RefCounted, add flag to clear whether or not this
instruction can commit. This is likely to be removed in the future.
cpu/beta_cpu/alpha_dyn_inst.cc:
AlphaDynInst has been changed to be templated, so now this CC file
is just used to force instantiations of AlphaDynInst.
cpu/beta_cpu/alpha_dyn_inst.hh:
Changed AlphaDynInst to be templated on Impl. Removed some unnecessary
functions.
cpu/beta_cpu/alpha_full_cpu.cc:
AlphaFullCPU has been changed to be templated, so this CC file is now
just used to force instantation of AlphaFullCPU.
cpu/beta_cpu/alpha_full_cpu.hh:
Change AlphaFullCPU to be templated on Impl.
cpu/beta_cpu/alpha_impl.hh:
Update it to reflect AlphaDynInst and AlphaFullCPU being templated
on Impl. Also removed time buffers from here, as they are really
a part of the CPU and are thus in the CPU policy now.
cpu/beta_cpu/alpha_params.hh:
Make AlphaSimpleParams inherit from the BaseFullCPU so that it doesn't
need to specifically declare any parameters that are already in the
BaseFullCPU.
cpu/beta_cpu/comm.hh:
Changed the structure of the time buffer communication structs. Now
they include the size of the packet of instructions it is sending.
Added some parameters to the backwards communication struct, mainly
for squashing.
cpu/beta_cpu/commit.hh:
Update typenames to reflect change in location of time buffer structs.
Update DynInst to DynInstPtr (it is refcounted now).
cpu/beta_cpu/commit_impl.hh:
Formatting changes mainly. Also sends back proper information
on branch mispredicts so that the bpred unit can update itself.
Updated behavior for non-speculative instructions (stores, any
other non-spec instructions): once they reach the head of the ROB,
the ROB signals back to the IQ that it can go ahead and issue the
non-speculative instruction. The instruction itself is updated so that
commit won't try to commit it again until it is done executing.
cpu/beta_cpu/cpu_policy.hh:
Added branch prediction unit, mem dependence prediction unit, load
store queue. Moved time buffer structs from AlphaSimpleImpl to here.
cpu/beta_cpu/decode.hh:
Changed typedefs to reflect change in location of time buffer structs
and also the change from DynInst to ref counted DynInstPtr.
cpu/beta_cpu/decode_impl.hh:
Continues to buffer instructions even while unblocking now. Changed
how it loops through groups of instructions so it can properly block
during the middle of a group of instructions.
cpu/beta_cpu/fetch.hh:
Changed typedefs to reflect change in location of time buffer structs
and the change to ref counted DynInsts. Also added in branch
brediction unit.
cpu/beta_cpu/fetch_impl.hh:
Add in branch prediction. Changed how fetch checks inputs and its
current state to make for easier logic.
cpu/beta_cpu/free_list.cc:
Changed int regs and float regs to logically use one flat namespace.
Future change will be moving them to a single scoreboard to conserve
space.
cpu/beta_cpu/free_list.hh:
Mostly debugging statements. Might be removed for performance in future.
cpu/beta_cpu/full_cpu.cc:
Added in some debugging statements. Updated BaseFullCPU to take
a params object.
cpu/beta_cpu/full_cpu.hh:
Added params class within BaseCPU that other param classes will be
able to inherit from. Updated typedefs to reflect change in location
of time buffer structs and ref counted DynInst.
cpu/beta_cpu/iew.hh:
Updated typedefs to reflect change in location of time buffer structs
and use of ref counted DynInsts.
cpu/beta_cpu/iew_impl.hh:
Added in load store queue, updated iew to be able to execute non-
speculative instructions, instead of having them execute in commit.
cpu/beta_cpu/inst_queue.hh:
Updated change to ref counted DynInsts. Changed inst queue to hold
non-speculative instructions as well, which are issued only when
commit signals backwards that a nonspeculative instruction is at
the head of the ROB.
cpu/beta_cpu/inst_queue_impl.hh:
Updated to allow for non-speculative instructions to be in the inst
queue. Also added some debug functions.
cpu/beta_cpu/regfile.hh:
Added debugging statements, changed formatting.
cpu/beta_cpu/rename.hh:
Updated typedefs, added some functions to clean up code.
cpu/beta_cpu/rename_impl.hh:
Moved some code into functions to make it easier to read.
cpu/beta_cpu/rename_map.cc:
Changed int and float reg behavior to use a single flat namespace. In
the future, the rename maps can be combined to a single rename map to
save space.
cpu/beta_cpu/rename_map.hh:
Added destructor.
cpu/beta_cpu/rob.hh:
Updated it with change from DynInst to ref counted DynInst.
cpu/beta_cpu/rob_impl.hh:
Formatting, updated to use ref counted DynInst.
cpu/static_inst.hh:
Updated forward declaration for AlphaDynInst now that it is templated.
--HG--
extra : convert_revision : 1045f240ee9b6a4bd368e1806aca029ebbdc6dd3
2004-09-23 20:06:03 +02:00
|
|
|
// If incorrect, then signal the ROB that it must be squashed.
|
2006-04-23 00:26:48 +02:00
|
|
|
squashDueToBranch(inst, tid);
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-12-16 13:32:06 +01:00
|
|
|
if (inst->readPredTaken()) {
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
predictedTakenIncorrect++;
|
2006-04-23 00:26:48 +02:00
|
|
|
} else {
|
|
|
|
predictedNotTakenIncorrect++;
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
}
|
2006-04-23 00:26:48 +02:00
|
|
|
} else if (ldstQueue.violation(tid)) {
|
2007-03-23 16:40:53 +01:00
|
|
|
assert(inst->isMemRef());
|
2006-05-19 21:44:03 +02:00
|
|
|
// If there was an ordering violation, then get the
|
|
|
|
// DynInst that caused the violation. Note that this
|
2006-04-23 00:26:48 +02:00
|
|
|
// clears the violation signal.
|
|
|
|
DynInstPtr violator;
|
|
|
|
violator = ldstQueue.getMemDepViolator(tid);
|
2004-08-20 20:54:07 +02:00
|
|
|
|
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
|
|
|
DPRINTF(IEW, "LDSTQ detected a violation. Violator PC: %s "
|
|
|
|
"[sn:%lli], inst PC: %s [sn:%lli]. Addr is: %#x.\n",
|
|
|
|
violator->pcState(), violator->seqNum,
|
|
|
|
inst->pcState(), inst->seqNum, inst->physEffAddr);
|
|
|
|
|
2006-12-12 23:35:46 +01:00
|
|
|
fetchRedirect[tid] = true;
|
|
|
|
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
// Tell the instruction queue that a violation has occured.
|
Update to make multiple instruction issue and different latencies work.
Also change to ref counted DynInst.
SConscript:
Add branch predictor, BTB, load store queue, and storesets.
arch/isa_parser.py:
Specify the template parameter for AlphaDynInst
base/traceflags.py:
Add load store queue, store set, and mem dependence unit to the
list of trace flags.
cpu/base_dyn_inst.cc:
Change formating, add in debug statement.
cpu/base_dyn_inst.hh:
Change DynInst to be RefCounted, add flag to clear whether or not this
instruction can commit. This is likely to be removed in the future.
cpu/beta_cpu/alpha_dyn_inst.cc:
AlphaDynInst has been changed to be templated, so now this CC file
is just used to force instantiations of AlphaDynInst.
cpu/beta_cpu/alpha_dyn_inst.hh:
Changed AlphaDynInst to be templated on Impl. Removed some unnecessary
functions.
cpu/beta_cpu/alpha_full_cpu.cc:
AlphaFullCPU has been changed to be templated, so this CC file is now
just used to force instantation of AlphaFullCPU.
cpu/beta_cpu/alpha_full_cpu.hh:
Change AlphaFullCPU to be templated on Impl.
cpu/beta_cpu/alpha_impl.hh:
Update it to reflect AlphaDynInst and AlphaFullCPU being templated
on Impl. Also removed time buffers from here, as they are really
a part of the CPU and are thus in the CPU policy now.
cpu/beta_cpu/alpha_params.hh:
Make AlphaSimpleParams inherit from the BaseFullCPU so that it doesn't
need to specifically declare any parameters that are already in the
BaseFullCPU.
cpu/beta_cpu/comm.hh:
Changed the structure of the time buffer communication structs. Now
they include the size of the packet of instructions it is sending.
Added some parameters to the backwards communication struct, mainly
for squashing.
cpu/beta_cpu/commit.hh:
Update typenames to reflect change in location of time buffer structs.
Update DynInst to DynInstPtr (it is refcounted now).
cpu/beta_cpu/commit_impl.hh:
Formatting changes mainly. Also sends back proper information
on branch mispredicts so that the bpred unit can update itself.
Updated behavior for non-speculative instructions (stores, any
other non-spec instructions): once they reach the head of the ROB,
the ROB signals back to the IQ that it can go ahead and issue the
non-speculative instruction. The instruction itself is updated so that
commit won't try to commit it again until it is done executing.
cpu/beta_cpu/cpu_policy.hh:
Added branch prediction unit, mem dependence prediction unit, load
store queue. Moved time buffer structs from AlphaSimpleImpl to here.
cpu/beta_cpu/decode.hh:
Changed typedefs to reflect change in location of time buffer structs
and also the change from DynInst to ref counted DynInstPtr.
cpu/beta_cpu/decode_impl.hh:
Continues to buffer instructions even while unblocking now. Changed
how it loops through groups of instructions so it can properly block
during the middle of a group of instructions.
cpu/beta_cpu/fetch.hh:
Changed typedefs to reflect change in location of time buffer structs
and the change to ref counted DynInsts. Also added in branch
brediction unit.
cpu/beta_cpu/fetch_impl.hh:
Add in branch prediction. Changed how fetch checks inputs and its
current state to make for easier logic.
cpu/beta_cpu/free_list.cc:
Changed int regs and float regs to logically use one flat namespace.
Future change will be moving them to a single scoreboard to conserve
space.
cpu/beta_cpu/free_list.hh:
Mostly debugging statements. Might be removed for performance in future.
cpu/beta_cpu/full_cpu.cc:
Added in some debugging statements. Updated BaseFullCPU to take
a params object.
cpu/beta_cpu/full_cpu.hh:
Added params class within BaseCPU that other param classes will be
able to inherit from. Updated typedefs to reflect change in location
of time buffer structs and ref counted DynInst.
cpu/beta_cpu/iew.hh:
Updated typedefs to reflect change in location of time buffer structs
and use of ref counted DynInsts.
cpu/beta_cpu/iew_impl.hh:
Added in load store queue, updated iew to be able to execute non-
speculative instructions, instead of having them execute in commit.
cpu/beta_cpu/inst_queue.hh:
Updated change to ref counted DynInsts. Changed inst queue to hold
non-speculative instructions as well, which are issued only when
commit signals backwards that a nonspeculative instruction is at
the head of the ROB.
cpu/beta_cpu/inst_queue_impl.hh:
Updated to allow for non-speculative instructions to be in the inst
queue. Also added some debug functions.
cpu/beta_cpu/regfile.hh:
Added debugging statements, changed formatting.
cpu/beta_cpu/rename.hh:
Updated typedefs, added some functions to clean up code.
cpu/beta_cpu/rename_impl.hh:
Moved some code into functions to make it easier to read.
cpu/beta_cpu/rename_map.cc:
Changed int and float reg behavior to use a single flat namespace. In
the future, the rename maps can be combined to a single rename map to
save space.
cpu/beta_cpu/rename_map.hh:
Added destructor.
cpu/beta_cpu/rob.hh:
Updated it with change from DynInst to ref counted DynInst.
cpu/beta_cpu/rob_impl.hh:
Formatting, updated to use ref counted DynInst.
cpu/static_inst.hh:
Updated forward declaration for AlphaDynInst now that it is templated.
--HG--
extra : convert_revision : 1045f240ee9b6a4bd368e1806aca029ebbdc6dd3
2004-09-23 20:06:03 +02:00
|
|
|
instQueue.violation(inst, violator);
|
|
|
|
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
// Squash.
|
2011-08-19 22:08:05 +02:00
|
|
|
squashDueToMemOrder(violator, tid);
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
|
|
|
++memOrderViolationEvents;
|
2006-04-23 00:26:48 +02:00
|
|
|
} else if (ldstQueue.loadBlocked(tid) &&
|
|
|
|
!ldstQueue.isLoadBlockedHandled(tid)) {
|
|
|
|
fetchRedirect[tid] = true;
|
|
|
|
|
|
|
|
DPRINTF(IEW, "Load operation couldn't execute because the "
|
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
|
|
|
"memory system is blocked. PC: %s [sn:%lli]\n",
|
|
|
|
inst->pcState(), inst->seqNum);
|
2006-04-23 00:26:48 +02:00
|
|
|
|
|
|
|
squashDueToMemBlocked(inst, tid);
|
Update to make multiple instruction issue and different latencies work.
Also change to ref counted DynInst.
SConscript:
Add branch predictor, BTB, load store queue, and storesets.
arch/isa_parser.py:
Specify the template parameter for AlphaDynInst
base/traceflags.py:
Add load store queue, store set, and mem dependence unit to the
list of trace flags.
cpu/base_dyn_inst.cc:
Change formating, add in debug statement.
cpu/base_dyn_inst.hh:
Change DynInst to be RefCounted, add flag to clear whether or not this
instruction can commit. This is likely to be removed in the future.
cpu/beta_cpu/alpha_dyn_inst.cc:
AlphaDynInst has been changed to be templated, so now this CC file
is just used to force instantiations of AlphaDynInst.
cpu/beta_cpu/alpha_dyn_inst.hh:
Changed AlphaDynInst to be templated on Impl. Removed some unnecessary
functions.
cpu/beta_cpu/alpha_full_cpu.cc:
AlphaFullCPU has been changed to be templated, so this CC file is now
just used to force instantation of AlphaFullCPU.
cpu/beta_cpu/alpha_full_cpu.hh:
Change AlphaFullCPU to be templated on Impl.
cpu/beta_cpu/alpha_impl.hh:
Update it to reflect AlphaDynInst and AlphaFullCPU being templated
on Impl. Also removed time buffers from here, as they are really
a part of the CPU and are thus in the CPU policy now.
cpu/beta_cpu/alpha_params.hh:
Make AlphaSimpleParams inherit from the BaseFullCPU so that it doesn't
need to specifically declare any parameters that are already in the
BaseFullCPU.
cpu/beta_cpu/comm.hh:
Changed the structure of the time buffer communication structs. Now
they include the size of the packet of instructions it is sending.
Added some parameters to the backwards communication struct, mainly
for squashing.
cpu/beta_cpu/commit.hh:
Update typenames to reflect change in location of time buffer structs.
Update DynInst to DynInstPtr (it is refcounted now).
cpu/beta_cpu/commit_impl.hh:
Formatting changes mainly. Also sends back proper information
on branch mispredicts so that the bpred unit can update itself.
Updated behavior for non-speculative instructions (stores, any
other non-spec instructions): once they reach the head of the ROB,
the ROB signals back to the IQ that it can go ahead and issue the
non-speculative instruction. The instruction itself is updated so that
commit won't try to commit it again until it is done executing.
cpu/beta_cpu/cpu_policy.hh:
Added branch prediction unit, mem dependence prediction unit, load
store queue. Moved time buffer structs from AlphaSimpleImpl to here.
cpu/beta_cpu/decode.hh:
Changed typedefs to reflect change in location of time buffer structs
and also the change from DynInst to ref counted DynInstPtr.
cpu/beta_cpu/decode_impl.hh:
Continues to buffer instructions even while unblocking now. Changed
how it loops through groups of instructions so it can properly block
during the middle of a group of instructions.
cpu/beta_cpu/fetch.hh:
Changed typedefs to reflect change in location of time buffer structs
and the change to ref counted DynInsts. Also added in branch
brediction unit.
cpu/beta_cpu/fetch_impl.hh:
Add in branch prediction. Changed how fetch checks inputs and its
current state to make for easier logic.
cpu/beta_cpu/free_list.cc:
Changed int regs and float regs to logically use one flat namespace.
Future change will be moving them to a single scoreboard to conserve
space.
cpu/beta_cpu/free_list.hh:
Mostly debugging statements. Might be removed for performance in future.
cpu/beta_cpu/full_cpu.cc:
Added in some debugging statements. Updated BaseFullCPU to take
a params object.
cpu/beta_cpu/full_cpu.hh:
Added params class within BaseCPU that other param classes will be
able to inherit from. Updated typedefs to reflect change in location
of time buffer structs and ref counted DynInst.
cpu/beta_cpu/iew.hh:
Updated typedefs to reflect change in location of time buffer structs
and use of ref counted DynInsts.
cpu/beta_cpu/iew_impl.hh:
Added in load store queue, updated iew to be able to execute non-
speculative instructions, instead of having them execute in commit.
cpu/beta_cpu/inst_queue.hh:
Updated change to ref counted DynInsts. Changed inst queue to hold
non-speculative instructions as well, which are issued only when
commit signals backwards that a nonspeculative instruction is at
the head of the ROB.
cpu/beta_cpu/inst_queue_impl.hh:
Updated to allow for non-speculative instructions to be in the inst
queue. Also added some debug functions.
cpu/beta_cpu/regfile.hh:
Added debugging statements, changed formatting.
cpu/beta_cpu/rename.hh:
Updated typedefs, added some functions to clean up code.
cpu/beta_cpu/rename_impl.hh:
Moved some code into functions to make it easier to read.
cpu/beta_cpu/rename_map.cc:
Changed int and float reg behavior to use a single flat namespace. In
the future, the rename maps can be combined to a single rename map to
save space.
cpu/beta_cpu/rename_map.hh:
Added destructor.
cpu/beta_cpu/rob.hh:
Updated it with change from DynInst to ref counted DynInst.
cpu/beta_cpu/rob_impl.hh:
Formatting, updated to use ref counted DynInst.
cpu/static_inst.hh:
Updated forward declaration for AlphaDynInst now that it is templated.
--HG--
extra : convert_revision : 1045f240ee9b6a4bd368e1806aca029ebbdc6dd3
2004-09-23 20:06:03 +02:00
|
|
|
}
|
2007-03-23 16:40:53 +01:00
|
|
|
} else {
|
|
|
|
// Reset any state associated with redirects that will not
|
|
|
|
// be used.
|
|
|
|
if (ldstQueue.violation(tid)) {
|
|
|
|
assert(inst->isMemRef());
|
|
|
|
|
|
|
|
DynInstPtr violator = ldstQueue.getMemDepViolator(tid);
|
|
|
|
|
|
|
|
DPRINTF(IEW, "LDSTQ detected a violation. Violator PC: "
|
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
|
|
|
"%s, inst PC: %s. Addr is: %#x.\n",
|
|
|
|
violator->pcState(), inst->pcState(),
|
|
|
|
inst->physEffAddr);
|
2007-03-23 16:40:53 +01:00
|
|
|
DPRINTF(IEW, "Violation will not be handled because "
|
|
|
|
"already squashing\n");
|
|
|
|
|
|
|
|
++memOrderViolationEvents;
|
|
|
|
}
|
|
|
|
if (ldstQueue.loadBlocked(tid) &&
|
|
|
|
!ldstQueue.isLoadBlockedHandled(tid)) {
|
|
|
|
DPRINTF(IEW, "Load operation couldn't execute because the "
|
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
|
|
|
"memory system is blocked. PC: %s [sn:%lli]\n",
|
|
|
|
inst->pcState(), inst->seqNum);
|
2007-03-23 16:40:53 +01:00
|
|
|
DPRINTF(IEW, "Blocked load will not be handled because "
|
|
|
|
"already squashing\n");
|
|
|
|
|
|
|
|
ldstQueue.setLoadBlockedHandled(tid);
|
|
|
|
}
|
|
|
|
|
2004-08-20 20:54:07 +02:00
|
|
|
}
|
|
|
|
}
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2006-05-31 17:45:02 +02:00
|
|
|
// Update and record activity if we processed any instructions.
|
2006-04-23 00:26:48 +02:00
|
|
|
if (inst_num) {
|
|
|
|
if (exeStatus == Idle) {
|
|
|
|
exeStatus = Running;
|
|
|
|
}
|
|
|
|
|
|
|
|
updatedQueues = true;
|
|
|
|
|
|
|
|
cpu->activityThisCycle();
|
|
|
|
}
|
|
|
|
|
|
|
|
// Need to reset this in case a writeback event needs to write into the
|
|
|
|
// iew queue. That way the writeback event will write into the correct
|
|
|
|
// spot in the queue.
|
|
|
|
wbNumInst = 0;
|
2011-01-18 23:30:05 +01:00
|
|
|
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
}
|
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
template <class Impl>
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
void
|
2006-04-23 00:26:48 +02:00
|
|
|
DefaultIEW<Impl>::writebackInsts()
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
{
|
2006-05-19 21:44:03 +02:00
|
|
|
// Loop through the head of the time buffer and wake any
|
|
|
|
// dependents. These instructions are about to write back. Also
|
|
|
|
// mark scoreboard that this instruction is finally complete.
|
|
|
|
// Either have IEW have direct access to scoreboard, or have this
|
|
|
|
// as part of backwards communication.
|
2006-12-19 00:20:13 +01:00
|
|
|
for (int inst_num = 0; inst_num < wbWidth &&
|
2006-04-23 00:26:48 +02:00
|
|
|
toCommit->insts[inst_num]; inst_num++) {
|
|
|
|
DynInstPtr inst = toCommit->insts[inst_num];
|
2009-05-26 18:23:13 +02:00
|
|
|
ThreadID tid = inst->threadNumber;
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
|
|
|
DPRINTF(IEW, "Sending instructions to commit, [sn:%lli] PC %s.\n",
|
|
|
|
inst->seqNum, inst->pcState());
|
2006-04-23 00:26:48 +02:00
|
|
|
|
2006-04-24 23:06:00 +02:00
|
|
|
iewInstsToCommit[tid]++;
|
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
// Some instructions will be sent to commit without having
|
|
|
|
// executed because they need commit to handle them.
|
|
|
|
// E.g. Uncached loads have not actually executed when they
|
|
|
|
// are first sent to commit. Instead commit must tell the LSQ
|
|
|
|
// when it's ready to execute the uncached load.
|
2006-09-28 06:09:27 +02:00
|
|
|
if (!inst->isSquashed() && inst->isExecuted() && inst->getFault() == NoFault) {
|
2006-04-24 23:06:00 +02:00
|
|
|
int dependents = instQueue.wakeDependents(inst);
|
2006-04-23 00:26:48 +02:00
|
|
|
|
|
|
|
for (int i = 0; i < inst->numDestRegs(); i++) {
|
|
|
|
//mark as Ready
|
|
|
|
DPRINTF(IEW,"Setting Destination Register %i\n",
|
|
|
|
inst->renamedDestRegIdx(i));
|
|
|
|
scoreboard->setReg(inst->renamedDestRegIdx(i));
|
|
|
|
}
|
2006-04-24 23:06:00 +02:00
|
|
|
|
2006-05-31 17:45:02 +02:00
|
|
|
if (dependents) {
|
|
|
|
producerInst[tid]++;
|
|
|
|
consumerInst[tid]+= dependents;
|
|
|
|
}
|
2006-05-19 21:44:03 +02:00
|
|
|
writebackCount[tid]++;
|
2006-04-23 00:26:48 +02:00
|
|
|
}
|
2006-07-05 21:51:36 +02:00
|
|
|
|
|
|
|
decrWb(inst->seqNum);
|
2006-04-23 00:26:48 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class Impl>
|
|
|
|
void
|
|
|
|
DefaultIEW<Impl>::tick()
|
|
|
|
{
|
|
|
|
wbNumInst = 0;
|
|
|
|
wbCycle = 0;
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
wroteToTimeBuffer = false;
|
|
|
|
updatedQueues = false;
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
sortInsts();
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-05-19 21:44:03 +02:00
|
|
|
// Free function units marked as being freed this cycle.
|
|
|
|
fuPool->processFreeUnits();
|
|
|
|
|
2009-05-26 18:23:13 +02:00
|
|
|
list<ThreadID>::iterator threads = activeThreads->begin();
|
|
|
|
list<ThreadID>::iterator end = activeThreads->end();
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-05-19 21:44:03 +02:00
|
|
|
// Check stall and squash signals, dispatch any instructions.
|
2006-12-21 07:20:11 +01:00
|
|
|
while (threads != end) {
|
2009-05-26 18:23:13 +02:00
|
|
|
ThreadID tid = *threads++;
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
DPRINTF(IEW,"Issue: Processing [tid:%i]\n",tid);
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
checkSignalsAndUpdate(tid);
|
|
|
|
dispatch(tid);
|
|
|
|
}
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
if (exeStatus != Squashing) {
|
|
|
|
executeInsts();
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
writebackInsts();
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
// Have the instruction queue try to schedule any ready instructions.
|
|
|
|
// (In actuality, this scheduling is for instructions that will
|
|
|
|
// be executed next cycle.)
|
|
|
|
instQueue.scheduleReadyInsts();
|
|
|
|
|
|
|
|
// Also should advance its own time buffers if the stage ran.
|
|
|
|
// Not the best place for it, but this works (hopefully).
|
|
|
|
issueToExecQueue.advance();
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
}
|
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
bool broadcast_free_entries = false;
|
|
|
|
|
|
|
|
if (updatedQueues || exeStatus == Running || updateLSQNextCycle) {
|
|
|
|
exeStatus = Idle;
|
|
|
|
updateLSQNextCycle = false;
|
|
|
|
|
|
|
|
broadcast_free_entries = true;
|
|
|
|
}
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
// Writeback any stores using any leftover bandwidth.
|
2005-05-03 16:56:47 +02:00
|
|
|
ldstQueue.writebackStores();
|
|
|
|
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
// Check the committed load/store signals to see if there's a load
|
|
|
|
// or store to commit. Also check if it's being told to execute a
|
|
|
|
// nonspeculative instruction.
|
2005-05-03 16:56:47 +02:00
|
|
|
// This is pretty inefficient...
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-12-21 07:20:11 +01:00
|
|
|
threads = activeThreads->begin();
|
|
|
|
while (threads != end) {
|
2009-05-26 18:23:13 +02:00
|
|
|
ThreadID tid = (*threads++);
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
DPRINTF(IEW,"Processing [tid:%i]\n",tid);
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-05-31 17:45:02 +02:00
|
|
|
// Update structures based on instructions committed.
|
2006-04-23 00:26:48 +02:00
|
|
|
if (fromCommit->commitInfo[tid].doneSeqNum != 0 &&
|
|
|
|
!fromCommit->commitInfo[tid].squash &&
|
|
|
|
!fromCommit->commitInfo[tid].robSquashing) {
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
ldstQueue.commitStores(fromCommit->commitInfo[tid].doneSeqNum,tid);
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
ldstQueue.commitLoads(fromCommit->commitInfo[tid].doneSeqNum,tid);
|
Check in of various updates to the CPU. Mainly adds in stats, improves
branch prediction, and makes memory dependence work properly.
SConscript:
Added return address stack, tournament predictor.
cpu/base_cpu.cc:
Added debug break and print statements.
cpu/base_dyn_inst.cc:
cpu/base_dyn_inst.hh:
Comment out possibly unneeded variables.
cpu/beta_cpu/2bit_local_pred.cc:
2bit predictor no longer speculatively updates itself.
cpu/beta_cpu/alpha_dyn_inst.hh:
Comment formatting.
cpu/beta_cpu/alpha_full_cpu.hh:
Formatting
cpu/beta_cpu/alpha_full_cpu_builder.cc:
Added new parameters for branch predictors, and IQ parameters.
cpu/beta_cpu/alpha_full_cpu_impl.hh:
Register stats.
cpu/beta_cpu/alpha_params.hh:
Added parameters for IQ, branch predictors, and store sets.
cpu/beta_cpu/bpred_unit.cc:
Removed one class.
cpu/beta_cpu/bpred_unit.hh:
Add in RAS, stats. Changed branch predictor unit functionality
so that it holds a history of past branches so it can update, and also
hold a proper history of the RAS so it can be restored on branch
mispredicts.
cpu/beta_cpu/bpred_unit_impl.hh:
Added in stats, history of branches, RAS. Now bpred unit actually
modifies the instruction's predicted next PC.
cpu/beta_cpu/btb.cc:
Add in sanity checks.
cpu/beta_cpu/comm.hh:
Add in communication where needed, remove it where it's not.
cpu/beta_cpu/commit.hh:
cpu/beta_cpu/rename.hh:
cpu/beta_cpu/rename_impl.hh:
Add in stats.
cpu/beta_cpu/commit_impl.hh:
Stats, update what is sent back on branch mispredict.
cpu/beta_cpu/cpu_policy.hh:
Change the bpred unit being used.
cpu/beta_cpu/decode.hh:
cpu/beta_cpu/decode_impl.hh:
Stats.
cpu/beta_cpu/fetch.hh:
Stats, change squash so it can handle squashes from decode differently
than squashes from commit.
cpu/beta_cpu/fetch_impl.hh:
Add in stats. Change how a cache line is fetched. Update to work with
caches. Also have separate functions for different behavior if squash
is coming from decode vs commit.
cpu/beta_cpu/free_list.hh:
Remove some old comments.
cpu/beta_cpu/full_cpu.cc:
cpu/beta_cpu/full_cpu.hh:
Added function to remove instructions from back of instruction list
until a certain sequence number.
cpu/beta_cpu/iew.hh:
Stats, separate squashing behavior due to branches vs memory.
cpu/beta_cpu/iew_impl.hh:
Stats, separate squashing behavior for branches vs memory.
cpu/beta_cpu/inst_queue.cc:
Debug stuff
cpu/beta_cpu/inst_queue.hh:
Stats, change how mem dep unit works, debug stuff
cpu/beta_cpu/inst_queue_impl.hh:
Stats, change how mem dep unit works, debug stuff. Also add in
parameters that used to be hardcoded.
cpu/beta_cpu/mem_dep_unit.hh:
cpu/beta_cpu/mem_dep_unit_impl.hh:
Add in stats, change how memory dependence unit works. It now holds
the memory instructions that are waiting for their memory dependences
to resolve. It provides which instructions are ready directly to the
IQ.
cpu/beta_cpu/regfile.hh:
Fix up sanity checks.
cpu/beta_cpu/rename_map.cc:
Fix loop variable type.
cpu/beta_cpu/rob_impl.hh:
Remove intermediate DynInstPtr
cpu/beta_cpu/store_set.cc:
Add in debugging statements.
cpu/beta_cpu/store_set.hh:
Reorder function arguments to match the rest of the calls.
--HG--
extra : convert_revision : aabf9b1fecd1d743265dfc3b174d6159937c6f44
2004-10-22 00:02:36 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
updateLSQNextCycle = true;
|
|
|
|
instQueue.commit(fromCommit->commitInfo[tid].doneSeqNum,tid);
|
|
|
|
}
|
2004-08-20 20:54:07 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
if (fromCommit->commitInfo[tid].nonSpecSeqNum != 0) {
|
2004-08-20 20:54:07 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
//DPRINTF(IEW,"NonspecInst from thread %i",tid);
|
|
|
|
if (fromCommit->commitInfo[tid].uncached) {
|
|
|
|
instQueue.replayMemInst(fromCommit->commitInfo[tid].uncachedLoad);
|
2007-03-23 16:40:53 +01:00
|
|
|
fromCommit->commitInfo[tid].uncachedLoad->setAtCommit();
|
2006-04-23 00:26:48 +02:00
|
|
|
} else {
|
|
|
|
instQueue.scheduleNonSpec(
|
|
|
|
fromCommit->commitInfo[tid].nonSpecSeqNum);
|
|
|
|
}
|
|
|
|
}
|
2004-08-20 20:54:07 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
if (broadcast_free_entries) {
|
|
|
|
toFetch->iewInfo[tid].iqCount =
|
|
|
|
instQueue.getCount(tid);
|
|
|
|
toFetch->iewInfo[tid].ldstqCount =
|
|
|
|
ldstQueue.getCount(tid);
|
2004-08-20 20:54:07 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
toRename->iewInfo[tid].usedIQ = true;
|
|
|
|
toRename->iewInfo[tid].freeIQEntries =
|
|
|
|
instQueue.numFreeEntries();
|
|
|
|
toRename->iewInfo[tid].usedLSQ = true;
|
|
|
|
toRename->iewInfo[tid].freeLSQEntries =
|
|
|
|
ldstQueue.numFreeEntries(tid);
|
|
|
|
|
|
|
|
wroteToTimeBuffer = true;
|
2004-08-20 20:54:07 +02:00
|
|
|
}
|
2006-04-23 00:26:48 +02:00
|
|
|
|
|
|
|
DPRINTF(IEW, "[tid:%i], Dispatch dispatched %i instructions.\n",
|
|
|
|
tid, toRename->iewInfo[tid].dispatched);
|
2004-08-20 20:54:07 +02:00
|
|
|
}
|
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
DPRINTF(IEW, "IQ has %i free entries (Can schedule: %i). "
|
|
|
|
"LSQ has %i free entries.\n",
|
|
|
|
instQueue.numFreeEntries(), instQueue.hasReadyInsts(),
|
|
|
|
ldstQueue.numFreeEntries());
|
2005-05-03 16:56:47 +02:00
|
|
|
|
2006-04-23 00:26:48 +02:00
|
|
|
updateStatus();
|
|
|
|
|
|
|
|
if (wroteToTimeBuffer) {
|
|
|
|
DPRINTF(Activity, "Activity this cycle.\n");
|
|
|
|
cpu->activityThisCycle();
|
|
|
|
}
|
2005-05-03 16:56:47 +02:00
|
|
|
}
|
2006-04-24 23:06:00 +02:00
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
void
|
|
|
|
DefaultIEW<Impl>::updateExeInstStats(DynInstPtr &inst)
|
|
|
|
{
|
2009-05-26 18:23:13 +02:00
|
|
|
ThreadID tid = inst->threadNumber;
|
2006-04-24 23:06:00 +02:00
|
|
|
|
|
|
|
//
|
|
|
|
// Pick off the software prefetches
|
|
|
|
//
|
|
|
|
#ifdef TARGET_ALPHA
|
|
|
|
if (inst->isDataPrefetch())
|
2009-05-26 18:23:13 +02:00
|
|
|
iewExecutedSwp[tid]++;
|
2006-04-24 23:06:00 +02:00
|
|
|
else
|
2006-06-14 04:35:05 +02:00
|
|
|
iewIewExecutedcutedInsts++;
|
2006-04-24 23:06:00 +02:00
|
|
|
#else
|
2006-06-03 00:15:20 +02:00
|
|
|
iewExecutedInsts++;
|
2006-04-24 23:06:00 +02:00
|
|
|
#endif
|
|
|
|
|
2011-07-15 18:53:35 +02:00
|
|
|
#if TRACING_ON
|
|
|
|
inst->completeTick = curTick();
|
|
|
|
#endif
|
|
|
|
|
2006-04-24 23:06:00 +02:00
|
|
|
//
|
|
|
|
// Control operations
|
|
|
|
//
|
|
|
|
if (inst->isControl())
|
2009-05-26 18:23:13 +02:00
|
|
|
iewExecutedBranches[tid]++;
|
2006-04-24 23:06:00 +02:00
|
|
|
|
|
|
|
//
|
|
|
|
// Memory operations
|
|
|
|
//
|
|
|
|
if (inst->isMemRef()) {
|
2009-05-26 18:23:13 +02:00
|
|
|
iewExecutedRefs[tid]++;
|
2006-04-24 23:06:00 +02:00
|
|
|
|
|
|
|
if (inst->isLoad()) {
|
2009-05-26 18:23:13 +02:00
|
|
|
iewExecLoadInsts[tid]++;
|
2006-04-24 23:06:00 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2010-08-23 18:18:40 +02:00
|
|
|
|
|
|
|
template <class Impl>
|
|
|
|
void
|
|
|
|
DefaultIEW<Impl>::checkMisprediction(DynInstPtr &inst)
|
|
|
|
{
|
|
|
|
ThreadID tid = inst->threadNumber;
|
|
|
|
|
|
|
|
if (!fetchRedirect[tid] ||
|
2011-01-18 23:30:05 +01:00
|
|
|
!toCommit->squash[tid] ||
|
2010-08-23 18:18:40 +02:00
|
|
|
toCommit->squashedSeqNum[tid] > inst->seqNum) {
|
|
|
|
|
|
|
|
if (inst->mispredicted()) {
|
|
|
|
fetchRedirect[tid] = true;
|
|
|
|
|
|
|
|
DPRINTF(IEW, "Execute: Branch mispredict detected.\n");
|
|
|
|
DPRINTF(IEW, "Predicted target was PC:%#x, NPC:%#x.\n",
|
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
|
|
|
inst->predInstAddr(), inst->predNextInstAddr());
|
2010-08-23 18:18:40 +02:00
|
|
|
DPRINTF(IEW, "Execute: Redirecting fetch to PC: %#x,"
|
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors.
This change is a low level and pervasive reorganization of how PCs are managed
in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about,
the PC and the NPC, and the lsb of the PC signaled whether or not you were in
PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next
micropc, x86 and ARM introduced variable length instruction sets, and ARM
started to keep track of mode bits in the PC. Each CPU model handled PCs in
its own custom way that needed to be updated individually to handle the new
dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack,
the complexity could be hidden in the ISA at the ISA implementation's expense.
Areas like the branch predictor hadn't been updated to handle branch delay
slots or micropcs, and it turns out that had introduced a significant (10s of
percent) performance bug in SPARC and to a lesser extend MIPS. Rather than
perpetuate the problem by reworking O3 again to handle the PC features needed
by x86, this change was introduced to rework PC handling in a more modular,
transparent, and hopefully efficient way.
PC type:
Rather than having the superset of all possible elements of PC state declared
in each of the CPU models, each ISA defines its own PCState type which has
exactly the elements it needs. A cross product of canned PCState classes are
defined in the new "generic" ISA directory for ISAs with/without delay slots
and microcode. These are either typedef-ed or subclassed by each ISA. To read
or write this structure through a *Context, you use the new pcState() accessor
which reads or writes depending on whether it has an argument. If you just
want the address of the current or next instruction or the current micro PC,
you can get those through read-only accessors on either the PCState type or
the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the
move away from readPC. That name is ambiguous since it's not clear whether or
not it should be the actual address to fetch from, or if it should have extra
bits in it like the PAL mode bit. Each class is free to define its own
functions to get at whatever values it needs however it needs to to be used in
ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the
PC and into a separate field like ARM.
These types can be reset to a particular pc (where npc = pc +
sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as
appropriate), printed, serialized, and compared. There is a branching()
function which encapsulates code in the CPU models that checked if an
instruction branched or not. Exactly what that means in the context of branch
delay slots which can skip an instruction when not taken is ambiguous, and
ideally this function and its uses can be eliminated. PCStates also generally
know how to advance themselves in various ways depending on if they point at
an instruction, a microop, or the last microop of a macroop. More on that
later.
Ideally, accessing all the PCs at once when setting them will improve
performance of M5 even though more data needs to be moved around. This is
because often all the PCs need to be manipulated together, and by getting them
all at once you avoid multiple function calls. Also, the PCs of a particular
thread will have spatial locality in the cache. Previously they were grouped
by element in arrays which spread out accesses.
Advancing the PC:
The PCs were previously managed entirely by the CPU which had to know about PC
semantics, try to figure out which dimension to increment the PC in, what to
set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction
with the PC type itself. Because most of the information about how to
increment the PC (mainly what type of instruction it refers to) is contained
in the instruction object, a new advancePC virtual function was added to the
StaticInst class. Subclasses provide an implementation that moves around the
right element of the PC with a minimal amount of decision making. In ISAs like
Alpha, the instructions always simply assign NPC to PC without having to worry
about micropcs, nnpcs, etc. The added cost of a virtual function call should
be outweighed by not having to figure out as much about what to do with the
PCs and mucking around with the extra elements.
One drawback of making the StaticInsts advance the PC is that you have to
actually have one to advance the PC. This would, superficially, seem to
require decoding an instruction before fetch could advance. This is, as far as
I can tell, realistic. fetch would advance through memory addresses, not PCs,
perhaps predicting new memory addresses using existing ones. More
sophisticated decisions about control flow would be made later on, after the
instruction was decoded, and handed back to fetch. If branching needs to
happen, some amount of decoding needs to happen to see that it's a branch,
what the target is, etc. This could get a little more complicated if that gets
done by the predecoder, but I'm choosing to ignore that for now.
Variable length instructions:
To handle variable length instructions in x86 and ARM, the predecoder now
takes in the current PC by reference to the getExtMachInst function. It can
modify the PC however it needs to (by setting NPC to be the PC + instruction
length, for instance). This could be improved since the CPU doesn't know if
the PC was modified and always has to write it back.
ISA parser:
To support the new API, all PC related operand types were removed from the
parser and replaced with a PCState type. There are two warts on this
implementation. First, as with all the other operand types, the PCState still
has to have a valid operand type even though it doesn't use it. Second, using
syntax like PCS.npc(target) doesn't work for two reasons, this looks like the
syntax for operand type overriding, and the parser can't figure out if you're
reading or writing. Instructions that use the PCS operand (which I've
consistently called it) need to first read it into a local variable,
manipulate it, and then write it back out.
Return address stack:
The return address stack needed a little extra help because, in the presence
of branch delay slots, it has to merge together elements of the return PC and
the call PC. To handle that, a buildRetPC utility function was added. There
are basically only two versions in all the ISAs, but it didn't seem short
enough to put into the generic ISA directory. Also, the branch predictor code
in O3 and InOrder were adjusted so that they always store the PC of the actual
call instruction in the RAS, not the next PC. If the call instruction is a
microop, the next PC refers to the next microop in the same macroop which is
probably not desirable. The buildRetPC function advances the PC intelligently
to the next macroop (in an ISA specific way) so that that case works.
Change in stats:
There were no change in stats except in MIPS and SPARC in the O3 model. MIPS
runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could
likely be improved further by setting call/return instruction flags and taking
advantage of the RAS.
TODO:
Add != operators to the PCState classes, defined trivially to be !(a==b).
Smooth out places where PCs are split apart, passed around, and put back
together later. I think this might happen in SPARC's fault code. Add ISA
specific constructors that allow setting PC elements without calling a bunch
of accessors. Try to eliminate the need for the branching() function. Factor
out Alpha's PAL mode pc bit into a separate flag field, and eliminate places
where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
|
|
|
" NPC: %#x.\n", inst->nextInstAddr(),
|
|
|
|
inst->nextInstAddr());
|
2010-08-23 18:18:40 +02:00
|
|
|
// If incorrect, then signal the ROB that it must be squashed.
|
|
|
|
squashDueToBranch(inst, tid);
|
|
|
|
|
|
|
|
if (inst->readPredTaken()) {
|
|
|
|
predictedTakenIncorrect++;
|
|
|
|
} else {
|
|
|
|
predictedNotTakenIncorrect++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|