ruby: Working M5 interface and updated Ruby interface.

This changeset also includes a lot of work from Derek Hower <drh5@cs.wisc.edu>

RubyMemory is now both a driver for Ruby and a port for M5.  Changed
makeRequest/hitCallback interface. Brought packets (superficially)
into the sequencer. Modified tester infrastructure to be packet based.
and Ruby can be used together through the example ruby_se.py
script. SPARC parallel applications work, and the timing *seems* right
from combined M5/Ruby debug traces. To run,
% build/ALPHA_SE/m5.debug configs/example/ruby_se.py -c
  tests/test-progs/hello/bin/alpha/linux/hello -n 4 -t
This commit is contained in:
Daniel Sanchez 2009-05-11 10:38:46 -07:00
parent ebf2f5aadd
commit 93f2f69657
29 changed files with 1149 additions and 301 deletions

190
configs/example/ruby.config Normal file
View file

@ -0,0 +1,190 @@
//Default parameters, taken from /athitos/export/08spr_ee382a/sanchezd/runs/gen-scripts/ruby.defaults
//General config
g_DEADLOCK_THRESHOLD: 20000000
RANDOMIZATION: false
g_tester_length: 0
SIMICS_RUBY_MULTIPLIER: 1
OPAL_RUBY_MULTIPLIER: 1
TRANSACTION_TRACE_ENABLED: false
USER_MODE_DATA_ONLY: false
PROFILE_HOT_LINES: false
PROFILE_ALL_INSTRUCTIONS: false
PRINT_INSTRUCTION_TRACE: false
g_DEBUG_CYCLE: 0
PERFECT_MEMORY_SYSTEM: false
PERFECT_MEMORY_SYSTEM_LATENCY: 0
DATA_BLOCK: false
// Line, page sizes
g_DATA_BLOCK_BYTES: 64
g_PAGE_SIZE_BYTES: 8192
g_REPLACEMENT_POLICY: PSEDUO_LRU
// For all caches (sic)
// L1 config
// 32KB, 4-way SA
L1_CACHE_ASSOC: 4
L1_CACHE_NUM_SETS_BITS: 7
// Single-cycle latency, hits take fastpath
SEQUENCER_TO_CONTROLLER_LATENCY: 1
REMOVE_SINGLE_CYCLE_DCACHE_FAST_PATH: false
// L1->L2 delays
L1_REQUEST_LATENCY: 1
L1_RESPONSE_LATENCY: 1
// L2 parameters
// 4 MB, 16-way SA
L2_CACHE_ASSOC: 16
L2_CACHE_NUM_SETS_BITS: 12
MAP_L2BANKS_TO_LOWEST_BITS: false
// Bank latencies
L2_RESPONSE_LATENCY: 10
L2_TAG_LATENCY: 5
// Directory latencies
// The one that counts, we have perfect dirs
DIRECTORY_CACHE_LATENCY: 6
// should not be used, but just in case...
DIRECTORY_LATENCY: 6
// Simple network parameters
// external links
NETWORK_LINK_LATENCY: 1
// intra-chip links
ON_CHIP_LINK_LATENCY: 1
// General latencies
RECYCLE_LATENCY: 1
//Used in MessageBuffer, also MSI_MOSI_CMP dir controller
// Unused parameters, good to define them to really weird things just in case
NULL_LATENCY: 100000
// Only SMP and token CMP protocols
ISSUE_LATENCY: 100000
// Only SMP, example protocols
CACHE_RESPONSE_LATENCY: 100000
// Only SMP protocols
COPY_HEAD_LATENCY: 100000
// In no protocols or ruby code
L2_RECYCLE_LATENCY: 100000
// In no protocols or ruby code
TIMER_LATENCY: 100000
// Not used
TBE_RESPONSE_LATENCY: 100000
// Not used
PERIODIC_TIMER_WAKEUPS: false
// Not used
BLOCK_STC: false
// Not used
SINGLE_ACCESS_L2_BANKS: false
// Not used
// Main memory latency
MEMORY_RESPONSE_LATENCY_MINUS_2: 448 //not used in _m, see below
PROFILE_EXCEPTIONS: false
PROFILE_XACT: false
PROFILE_NONXACT: true
XACT_DEBUG: false
XACT_DEBUG_LEVEL: 1
XACT_MEMORY: false
XACT_ENABLE_TOURMALINE: false
XACT_NUM_CURRENT: 0
XACT_LAST_UPDATE: 0
XACT_ISOLATION_CHECK: false
PERFECT_FILTER: true
READ_WRITE_FILTER: Perfect_
PERFECT_VIRTUAL_FILTER: true
VIRTUAL_READ_WRITE_FILTER: Perfect_
PERFECT_SUMMARY_FILTER: true
SUMMARY_READ_WRITE_FILTER: Perfect_
XACT_EAGER_CD: true
XACT_LAZY_VM: false
XACT_CONFLICT_RES: BASE
XACT_COMMIT_TOKEN_LATENCY: 0
XACT_NO_BACKOFF: false
XACT_LOG_BUFFER_SIZE: 0
XACT_STORE_PREDICTOR_HISTORY: 0
XACT_STORE_PREDICTOR_ENTRIES: 0
XACT_STORE_PREDICTOR_THRESHOLD: 0
XACT_FIRST_ACCESS_COST: 0
XACT_FIRST_PAGE_ACCESS_COST: 0
ENABLE_MAGIC_WAITING: false
ENABLE_WATCHPOINT: false
XACT_ENABLE_VIRTUALIZATION_LOGTM_SE: false
ATMTP_ENABLED: false
ATMTP_ABORT_ON_NON_XACT_INST: false
ATMTP_ALLOW_SAVE_RESTORE_IN_XACT: false
ATMTP_XACT_MAX_STORES: 0
ATMTP_DEBUG_LEVEL: 0
XACT_LENGTH: 0
XACT_SIZE: 0
ABORT_RETRY_TIME: 0
// Allowed parallelism in controllers
L1CACHE_TRANSITIONS_PER_RUBY_CYCLE: 32
L2CACHE_TRANSITIONS_PER_RUBY_CYCLE: 1000
DIRECTORY_TRANSITIONS_PER_RUBY_CYCLE: 1000
g_SEQUENCER_OUTSTANDING_REQUESTS: 16
//TBEs == MSHRs (global)
NUMBER_OF_TBES: 128
NUMBER_OF_L1_TBES: 32
// unused in CMP protocols
NUMBER_OF_L2_TBES: 32
// unused in CMP protocols
// TSO & WBuffer params (unused)
FINITE_BUFFERING: false
FINITE_BUFFER_SIZE: 3
PROCESSOR_BUFFER_SIZE: 10
PROTOCOL_BUFFER_SIZE: 32
TSO: false
// General network params
g_endpoint_bandwidth: 10000
g_adaptive_routing: true
NUMBER_OF_VIRTUAL_NETWORKS: 5
FAN_OUT_DEGREE: 4
// for HIERARCHICAL_SWITCH
// Detailed Memory Controller Params (only used in _m protocols)
MEM_BUS_CYCLE_MULTIPLIER: 5
BANKS_PER_RANK: 8
RANKS_PER_DIMM: 2
DIMMS_PER_CHANNEL: 2
BANK_BIT_0: 8
RANK_BIT_0: 11
DIMM_BIT_0: 12
BANK_QUEUE_SIZE: 12
BANK_BUSY_TIME: 22
RANK_RANK_DELAY: 2
READ_WRITE_DELAY: 3
BASIC_BUS_BUSY_TIME: 3
MEM_CTL_LATENCY: 20
REFRESH_PERIOD: 3120
TFAW: 0
//flip a coin to delay requests by one cycle, introduces non-determinism
MEM_RANDOM_ARBITRATE: 50
MEM_FIXED_DELAY: 0
//Configuration-specific parameters
g_NUM_PROCESSORS: 1
g_NUM_CHIPS: 1
g_PROCS_PER_CHIP: 1
g_NUM_L2_BANKS: 1
g_NUM_MEMORIES: 4
g_PRINT_TOPOLOGY: true
g_GARNET_NETWORK: true
g_DETAIL_NETWORK: true
g_FLIT_SIZE: 8

172
configs/example/ruby_se.py Normal file
View file

@ -0,0 +1,172 @@
# Copyright (c) 2006-2008 The Regents of The University of Michigan
# All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are
# met: redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer;
# redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in the
# documentation and/or other materials provided with the distribution;
# neither the name of the copyright holders nor the names of its
# contributors may be used to endorse or promote products derived from
# this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#
# Authors: Steve Reinhardt
# Simple test script
#
# "m5 test.py"
import m5
if m5.build_env['FULL_SYSTEM']:
m5.panic("This script requires syscall emulation mode (*_SE).")
from m5.objects import *
import os, optparse, sys
from os.path import join as joinpath
m5.AddToPath('../common')
import Simulation
#from Caches import *
from cpu2000 import *
# Get paths we might need. It's expected this file is in m5/configs/example.
config_path = os.path.dirname(os.path.abspath(__file__))
config_root = os.path.dirname(config_path)
m5_root = os.path.dirname(config_root)
parser = optparse.OptionParser()
# Benchmark options
parser.add_option("-c", "--cmd",
default=joinpath(m5_root, "tests/test-progs/hello/bin/alpha/linux/hello"),
help="The binary to run in syscall emulation mode.")
parser.add_option("-o", "--options", default="",
help='The options to pass to the binary, use " " around the entire string')
parser.add_option("-i", "--input", default="", help="Read stdin from a file.")
parser.add_option("--output", default="", help="Redirect stdout to a file.")
parser.add_option("--errout", default="", help="Redirect stderr to a file.")
parser.add_option("--ruby-debug", action="store_true")
parser.add_option("--ruby-debug-file", default="", help="Ruby debug out file (stdout if blank)")
execfile(os.path.join(config_root, "common", "Options.py"))
(options, args) = parser.parse_args()
if args:
print "Error: script doesn't take any positional arguments"
sys.exit(1)
if options.bench:
try:
if m5.build_env['TARGET_ISA'] != 'alpha':
print >>sys.stderr, "Simpoints code only works for Alpha ISA at this time"
sys.exit(1)
exec("workload = %s('alpha', 'tru64', 'ref')" % options.bench)
process = workload.makeLiveProcess()
except:
print >>sys.stderr, "Unable to find workload for %s" % options.bench
sys.exit(1)
else:
process = LiveProcess()
process.executable = options.cmd
process.cmd = [options.cmd] + options.options.split()
if options.input != "":
process.input = options.input
if options.output != "":
process.output = options.output
if options.errout != "":
process.errout = options.errout
if options.detailed:
#check for SMT workload
workloads = options.cmd.split(';')
if len(workloads) > 1:
process = []
smt_idx = 0
inputs = []
outputs = []
errouts = []
if options.input != "":
inputs = options.input.split(';')
if options.output != "":
outputs = options.output.split(';')
if options.errout != "":
errouts = options.errout.split(';')
for wrkld in workloads:
smt_process = LiveProcess()
smt_process.executable = wrkld
smt_process.cmd = wrkld + " " + options.options
if inputs and inputs[smt_idx]:
smt_process.input = inputs[smt_idx]
if outputs and outputs[smt_idx]:
smt_process.output = outputs[smt_idx]
if errouts and errouts[smt_idx]:
smt_process.errout = errouts[smt_idx]
process += [smt_process, ]
smt_idx += 1
(CPUClass, test_mem_mode, FutureClass) = Simulation.setCPUClass(options)
CPUClass.clock = '1GHz'
np = options.num_cpus
rubymem = RubyMemory(
range = AddrRange("512MB"),
clock = "1GHz",
num_cpus = np,
libruby_file = "src/mem/ruby/amd64-linux/generated/MOESI_CMP_directory/bin/libruby.so",
config_file = "ruby.config",
stats_file = "m5out/ruby.stats"
)
if options.ruby_debug == True:
rubymem.debug = True
rubymem.debug_file = options.ruby_debug_file
system = System(cpu = [CPUClass(cpu_id=i) for i in xrange(np)],
physmem = rubymem)
if options.l2cache:
print "Error: -l2cache incompatible with ruby, must configure it ruby-style"
sys.exit(1)
if options.caches:
print "Error: -caches incompatible with ruby, must configure it ruby-style"
sys.exit(1)
for i in xrange(np):
system.cpu[i].connectMemPorts(system.physmem)
'''process = LiveProcess()
process.executable = options.cmd
process.cmd = [options.cmd, str(i)]
'''
system.cpu[i].workload = process
if options.fastmem:
system.cpu[i].physmem_port = system.physmem.port
root = Root(system = system)
Simulation.run(options, root, system, FutureClass)

46
src/mem/RubyMemory.py Normal file
View file

@ -0,0 +1,46 @@
# Copyright (c) 2005-2008 The Regents of The University of Michigan
# All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are
# met: redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer;
# redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in the
# documentation and/or other materials provided with the distribution;
# neither the name of the copyright holders nor the names of its
# contributors may be used to endorse or promote products derived from
# this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#
# Authors: Nathan Binkert
from m5.params import *
from m5.proxy import *
from PhysicalMemory import PhysicalMemory
class RubyMemory(PhysicalMemory):
type = 'RubyMemory'
clock = Param.Clock('1t', "ruby clock speed")
phase = Param.Latency('0ns', "ruby clock phase")
config_file = Param.String("", "path to the Ruby config file")
config_options = Param.String("", "extra Ruby options (one per line)")
stats_file = Param.String("ruby.stats",
"file to which ruby dumps its stats")
num_cpus = Param.Int(1, "Number of CPUs connected to the Ruby memory")
debug = Param.Bool(False, "Use ruby debug")
debug_file = Param.String("",
"path to the Ruby debug output file (stdout if blank)")

View file

@ -32,8 +32,9 @@ Import('*')
SimObject('Bridge.py') SimObject('Bridge.py')
SimObject('Bus.py') SimObject('Bus.py')
SimObject('PhysicalMemory.py')
SimObject('MemObject.py') SimObject('MemObject.py')
SimObject('PhysicalMemory.py')
SimObject('RubyMemory.py')
Source('bridge.cc') Source('bridge.cc')
Source('bus.cc') Source('bus.cc')
@ -44,6 +45,7 @@ Source('physical.cc')
Source('port.cc') Source('port.cc')
Source('tport.cc') Source('tport.cc')
Source('mport.cc') Source('mport.cc')
Source('rubymem.cc')
if env['FULL_SYSTEM']: if env['FULL_SYSTEM']:
Source('vport.cc') Source('vport.cc')

View file

@ -64,8 +64,6 @@ public:
* @param initializingString A string (with value pairs) for initialization * @param initializingString A string (with value pairs) for initialization
* @param allocate_f A ptr to the allocate function * @param allocate_f A ptr to the allocate function
* @param generate_values A ptr to the generate values function * @param generate_values A ptr to the generate values function
* @param my_get_attr A ptr to the get attribute function
* @param my_set_attr A ptr to the set attribute function
*/ */
initvar_t( const char *name, const char *relativeIncludePath, initvar_t( const char *name, const char *relativeIncludePath,
const char *initializingString, const char *initializingString,

View file

@ -34,6 +34,7 @@
#include "mem/ruby/common/Consumer.hh" #include "mem/ruby/common/Consumer.hh"
#include "mem/ruby/system/NodeID.hh" #include "mem/ruby/system/NodeID.hh"
#include "mem/protocol/CacheRequestType.hh" #include "mem/protocol/CacheRequestType.hh"
#include "mem/packet.hh"
class RubySystem; class RubySystem;
class SubBlock; class SubBlock;
@ -51,27 +52,10 @@ public:
// Public Methods // Public Methods
virtual void get_network_config() {} virtual void get_network_config() {}
virtual void hitCallback(NodeID proc, SubBlock& data, CacheRequestType type, int thread) = 0; // Called by sequencer virtual void hitCallback(Packet* pkt) = 0;
virtual void conflictCallback(NodeID proc, SubBlock& data, CacheRequestType type, int thread) { assert(0); } // Called by sequencer
virtual integer_t getInstructionCount(int procID) const { return 1; } virtual integer_t getInstructionCount(int procID) const { return 1; }
virtual integer_t getCycleCount(int procID) const { return 1; } virtual integer_t getCycleCount(int procID) const { return 1; }
virtual SimicsHypervisor * getHypervisor() { return NULL; }
virtual void notifySendNack( int procID, const Address & addr, uint64 remote_timestamp, const MachineID & remote_id) { assert(0); }; //Called by Sequencer
virtual void notifyReceiveNack( int procID, const Address & addr, uint64 remote_timestamp, const MachineID & remote_id) { assert(0); }; //Called by Sequencer
virtual void notifyReceiveNackFinal( int procID, const Address & addr) { assert(0); }; // Called by Sequencer
virtual void notifyTrapStart( int procID, const Address & handlerPC, int threadID, int smtThread ) { assert(0); } //called by Sequencer
virtual void notifyTrapComplete( int procID, const Address & newPC, int smtThread ) {assert(0); } // called by Sequencer
virtual int getOpalTransactionLevel(int procID, int thread) const {
cout << "mem/ruby/common/Driver.hh getOpalTransactionLevel() " << endl;
return 0; } //called by Sequencer
virtual void addThreadDependency(int procID, int requestor_thread, int conflict_thread) const { assert(0);}
virtual uint64 getOpalTime(int procID) const{ return 0; } //called by Sequencer
virtual uint64 getOpalTimestamp(int procID, int thread) const{
cout << "mem/ruby/common/Driver.hh getOpalTimestamp " << endl;
return 0; } // called by Sequencer
virtual int inTransaction(int procID, int thread ) const{
cout << "mem/ruby/common/Driver.hh inTransaction " << endl;
return false; } //called by Sequencer
virtual void printDebug(){} //called by Sequencer virtual void printDebug(){} //called by Sequencer
virtual void printStats(ostream& out) const = 0; virtual void printStats(ostream& out) const = 0;
@ -79,8 +63,6 @@ return false; } //called by Sequencer
virtual void printConfig(ostream& out) const = 0; virtual void printConfig(ostream& out) const = 0;
//virtual void abortCallback(NodeID proc){}
virtual integer_t readPhysicalMemory(int procID, physical_address_t address, virtual integer_t readPhysicalMemory(int procID, physical_address_t address,
int len ){ ASSERT(0); return 0; } int len ){ ASSERT(0); return 0; }

View file

@ -40,6 +40,7 @@
#include "mem/ruby/eventqueue/RubyEventQueue.hh" #include "mem/ruby/eventqueue/RubyEventQueue.hh"
#include "mem/ruby/system/System.hh" #include "mem/ruby/system/System.hh"
#include "mem/ruby/common/Debug.hh" #include "mem/ruby/common/Debug.hh"
#include "mem/ruby/common/Driver.hh"
#include "mem/ruby/profiler/Profiler.hh" #include "mem/ruby/profiler/Profiler.hh"
#include "mem/ruby/tester/Tester.hh" #include "mem/ruby/tester/Tester.hh"
#include "mem/ruby/init.hh" #include "mem/ruby/init.hh"
@ -65,14 +66,27 @@ static void init_generate_values( void )
} }
//*************************************************************************** //***************************************************************************
void init_variables( void ) void init_variables( void )
{ {
// allocate the "variable initialization" package // allocate the "variable initialization" package
ruby_initvar_obj = new initvar_t( "ruby", "../../../ruby/", ruby_initvar_obj = new initvar_t( "ruby", "../../../ruby/",
global_default_param, global_default_param,
&init_simulator, &init_simulator,
&init_generate_values);
}
/*
void init_variables(const char* config_str )
{
// allocate the "variable initialization" package
ruby_initvar_obj = new initvar_t( "ruby", "../../../ruby/",
config_str,
&init_simulator,
&init_generate_values ); &init_generate_values );
} }
*/
void init_simulator() void init_simulator()
{ {
@ -104,6 +118,36 @@ void init_simulator()
cout << "Ruby initialization complete" << endl; cout << "Ruby initialization complete" << endl;
} }
void init_simulator(Driver* _driver)
{
// Set things to NULL to make sure we don't de-reference them
// without a seg. fault.
g_system_ptr = NULL;
g_debug_ptr = NULL;
g_eventQueue_ptr = NULL;
cout << "Ruby Timing Mode" << endl;
g_debug_ptr = new Debug( DEBUG_FILTER_STRING,
DEBUG_VERBOSITY_STRING,
DEBUG_START_TIME,
DEBUG_OUTPUT_FILENAME );
RubyConfig::init();
cout << "Creating event queue..." << endl;
g_eventQueue_ptr = new RubyEventQueue;
cout << "Creating event queue done" << endl;
cout << "Creating system..." << endl;
cout << " Processors: " << RubyConfig::numberOfProcessors() << endl;
g_system_ptr = new RubySystem(_driver);
cout << "Creating system done" << endl;
cout << "Ruby initialization complete" << endl;
}
void destroy_simulator() void destroy_simulator()
{ {
cout << "Deleting system..." << endl; cout << "Deleting system..." << endl;
@ -122,7 +166,8 @@ void destroy_simulator()
| M5 in phase 1 integration, and possibly afterwards, too. | | M5 in phase 1 integration, and possibly afterwards, too. |
+-------------------------------------------------------------------------*/ +-------------------------------------------------------------------------*/
extern "C" //dsm: superfluous
/*extern "C"
int OnLoadRuby() { int OnLoadRuby() {
init_variables(); init_variables();
return 0; return 0;
@ -138,7 +183,7 @@ extern "C"
int OnUnloadRuby() { int OnUnloadRuby() {
destroy_simulator(); destroy_simulator();
return 0; return 0;
} }*/
/* I have to put it somewhere for now */ /* I have to put it somewhere for now */
void tester_main(int argc, char **argv) { void tester_main(int argc, char **argv) {

View file

@ -39,8 +39,12 @@
#ifndef INIT_H #ifndef INIT_H
#define INIT_H #define INIT_H
class Driver;
extern void init_variables(); extern void init_variables();
//extern void init_variables(const char* config_str);
extern void init_simulator(); extern void init_simulator();
extern void init_simulator(Driver* _driver);
extern void destroy_simulator(); extern void destroy_simulator();
#endif //INIT_H #endif //INIT_H

View file

@ -1,165 +0,0 @@
/*
* Copyright (c) 1999-2008 Mark D. Hill and David A. Wood
* 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.
*/
/*------------------------------------------------------------------------*/
/* Includes */
/*------------------------------------------------------------------------*/
/*------------------------------------------------------------------------*/
/* Macro declarations */
/*------------------------------------------------------------------------*/
#ifndef _MF_MEMORY_API_H_
#define _MF_MEMORY_API_H_
#ifdef SIMICS30
#ifndef pa_t
typedef physical_address_t pa_t;
typedef physical_address_t la_t;
#endif
#endif
/**
* Defines types of memory requests
*/
typedef enum OpalMemop {
OPAL_LOAD,
OPAL_STORE,
OPAL_IFETCH,
OPAL_ATOMIC,
} OpalMemop_t;
/*------------------------------------------------------------------------*/
/* Class declaration(s) */
/*------------------------------------------------------------------------*/
/**
* structure which provides an interface between ruby and opal.
*/
typedef struct mf_opal_api {
/**
* @name Methods
*/
//@{
/**
* notify processor model that data from address address is available at proc
*/
void (*hitCallback)( int cpuNumber, pa_t phys_address, OpalMemop_t type, int thread );
/**
* notify opal that ruby is loaded, or removed
*/
void (*notifyCallback)( int status );
/**
* query for the number of instructions executed on a given processor.
*/
integer_t (*getInstructionCount)( int cpuNumber );
// for printing out debug info on crash
void (*printDebug)();
/** query Opal for the current time */
uint64 (*getOpalTime)(int cpuNumber);
/** For WATTCH power stats */
// Called whenever L2 is accessed
void (*incrementL2Access)(int cpuNumber);
// Called whenever prefetcher is accessed
void (*incrementPrefetcherAccess)(int cpuNumber, int num_prefetches, int isinstr);
/* Called whenever there's an L2 miss */
void (*notifyL2Miss)(int cpuNumber, physical_address_t physicalAddr, OpalMemop_t type, int tagexists);
//@}
} mf_opal_api_t;
typedef struct mf_ruby_api {
/**
* @name Methods
*/
//@{
/**
* Check to see if the system is ready for more requests
*/
int (*isReady)( int cpuNumber, la_t logicalAddr, pa_t physicalAddr, OpalMemop_t typeOfRequest, int thread );
/**
* Make a 'mandatory' request to the memory hierarchy
*/
void (*makeRequest)( int cpuNumber, la_t logicalAddr, pa_t physicalAddr,
int requestSize, OpalMemop_t typeOfRequest,
la_t virtualPC, int isPriv, int thread);
/**
* Make a prefetch request to the memory hierarchy
*/
void (*makePrefetch)( int cpuNumber, la_t logicalAddr, pa_t physicalAddr,
int requestSize, OpalMemop_t typeOfRequest,
la_t virtualPC, int isPriv, int thread);
/**
* Ask the memory hierarchy for 'stale' data that can be used for speculation
* Returns true (1) if the tag matches, false (0) if not.
*/
int (*staleDataRequest)( int cpuNumber, pa_t physicalAddr,
int requestSize, int8 *buffer );
/**
* Advance ruby's cycle time one step
*/
void (*advanceTime)( void );
/**
* Get ruby's cycle time count.
*/
uint64 (*getTime)( void );
/** prints Ruby's outstanding request table */
void (*printProgress)(int cpuNumber);
/**
* notify ruby that opal is loaded, or removed
*/
void (*notifyCallback)( int status );
// Returns the number of outstanding request
int (*getNumberOutstanding)(int cpuNumber);
// Returns the number of outstanding demand requests
int (*getNumberOutstandingDemand)(int cpuNumber );
// Returns the number of outstanding prefetch request
int (*getNumberOutstandingPrefetch)(int cpuNumber );
//@}
} mf_ruby_api_t;
#endif //_MF_MEMORY_API_H_

View file

@ -37,6 +37,7 @@
#include "mem/ruby/system/System.hh" #include "mem/ruby/system/System.hh"
#include "mem/ruby/slicc_interface/AbstractChip.hh" #include "mem/ruby/slicc_interface/AbstractChip.hh"
#include "mem/protocol/CacheMsg.hh" #include "mem/protocol/CacheMsg.hh"
#include "mem/packet.hh"
TraceRecord::TraceRecord(NodeID id, const Address& data_addr, const Address& pc_addr, CacheRequestType type, Time time) TraceRecord::TraceRecord(NodeID id, const Address& data_addr, const Address& pc_addr, CacheRequestType type, Time time)
{ {
@ -78,14 +79,27 @@ void TraceRecord::issueRequest() const
Sequencer* sequencer_ptr = chip_ptr->getSequencer((m_node_num/RubyConfig::numberofSMTThreads())%RubyConfig::numberOfProcsPerChip()); Sequencer* sequencer_ptr = chip_ptr->getSequencer((m_node_num/RubyConfig::numberofSMTThreads())%RubyConfig::numberOfProcsPerChip());
assert(sequencer_ptr != NULL); assert(sequencer_ptr != NULL);
CacheMsg request(m_data_address, m_data_address, m_type, m_pc_address, AccessModeType_UserMode, 0, PrefetchBit_Yes, 0, Address(0), 0 /* only 1 SMT thread */); Addr data_addr = m_data_address.getAddress();
Addr pc_addr = m_pc_address.getAddress();
Request request(0, data_addr, 0, Flags<unsigned int>(Request::PREFETCH), pc_addr, m_node_num, 0);
MemCmd::Command command;
if (m_type == CacheRequestType_LD || m_type == CacheRequestType_IFETCH)
command = MemCmd::ReadReq;
else if (m_type == CacheRequestType_ST)
command = MemCmd::WriteReq;
else if (m_type == CacheRequestType_ATOMIC)
command = MemCmd::SwapReq; // TODO -- differentiate between atomic types
else
assert(false);
Packet pkt(&request, command, 0); // TODO -- make dest a real NodeID
// Clear out the sequencer // Clear out the sequencer
while (!sequencer_ptr->empty()) { while (!sequencer_ptr->empty()) {
g_eventQueue_ptr->triggerEvents(g_eventQueue_ptr->getTime() + 100); g_eventQueue_ptr->triggerEvents(g_eventQueue_ptr->getTime() + 100);
} }
sequencer_ptr->makeRequest(request); sequencer_ptr->makeRequest(&pkt);
// Clear out the sequencer // Clear out the sequencer
while (!sequencer_ptr->empty()) { while (!sequencer_ptr->empty()) {

View file

@ -42,7 +42,9 @@
#include "mem/ruby/common/Global.hh" #include "mem/ruby/common/Global.hh"
#include "mem/protocol/AccessPermission.hh" #include "mem/protocol/AccessPermission.hh"
#include "mem/ruby/common/Address.hh" #include "mem/ruby/common/Address.hh"
#include "mem/ruby/recorder/CacheRecorder.hh"
//dsm: PRUNED
//#include "mem/ruby/recorder/CacheRecorder.hh"
#include "mem/protocol/CacheRequestType.hh" #include "mem/protocol/CacheRequestType.hh"
#include "mem/gems_common/Vector.hh" #include "mem/gems_common/Vector.hh"
#include "mem/ruby/common/DataBlock.hh" #include "mem/ruby/common/DataBlock.hh"
@ -142,6 +144,8 @@ private:
int m_cache_num_sets; int m_cache_num_sets;
int m_cache_num_set_bits; int m_cache_num_set_bits;
int m_cache_assoc; int m_cache_assoc;
bool is_locked; // for LL/SC
}; };
// Output operator declaration // Output operator declaration
@ -489,7 +493,10 @@ template<class ENTRY>
inline inline
void CacheMemory<ENTRY>::recordCacheContents(CacheRecorder& tr) const void CacheMemory<ENTRY>::recordCacheContents(CacheRecorder& tr) const
{ {
for (int i = 0; i < m_cache_num_sets; i++) { //dsm: Uses CacheRecorder, PRUNED
assert(false);
/* for (int i = 0; i < m_cache_num_sets; i++) {
for (int j = 0; j < m_cache_assoc; j++) { for (int j = 0; j < m_cache_assoc; j++) {
AccessPermission perm = m_cache[i][j].m_Permission; AccessPermission perm = m_cache[i][j].m_Permission;
CacheRequestType request_type = CacheRequestType_NULL; CacheRequestType request_type = CacheRequestType_NULL;
@ -508,7 +515,7 @@ void CacheMemory<ENTRY>::recordCacheContents(CacheRecorder& tr) const
Address(0), request_type, m_replacementPolicy_ptr->getLastAccess(i, j)); Address(0), request_type, m_replacementPolicy_ptr->getLastAccess(i, j));
} }
} }
} }*/
} }
template<class ENTRY> template<class ENTRY>

View file

@ -46,6 +46,7 @@
#include "mem/ruby/common/SubBlock.hh" #include "mem/ruby/common/SubBlock.hh"
#include "mem/protocol/Protocol.hh" #include "mem/protocol/Protocol.hh"
#include "mem/gems_common/Map.hh" #include "mem/gems_common/Map.hh"
#include "mem/packet.hh"
Sequencer::Sequencer(AbstractChip* chip_ptr, int version) { Sequencer::Sequencer(AbstractChip* chip_ptr, int version) {
m_chip_ptr = chip_ptr; m_chip_ptr = chip_ptr;
@ -58,6 +59,8 @@ Sequencer::Sequencer(AbstractChip* chip_ptr, int version) {
m_writeRequestTable_ptr = new Map<Address, CacheMsg>*[smt_threads]; m_writeRequestTable_ptr = new Map<Address, CacheMsg>*[smt_threads];
m_readRequestTable_ptr = new Map<Address, CacheMsg>*[smt_threads]; m_readRequestTable_ptr = new Map<Address, CacheMsg>*[smt_threads];
m_packetTable_ptr = new Map<Address, Packet*>;
for(int p=0; p < smt_threads; ++p){ for(int p=0; p < smt_threads; ++p){
m_writeRequestTable_ptr[p] = new Map<Address, CacheMsg>; m_writeRequestTable_ptr[p] = new Map<Address, CacheMsg>;
m_readRequestTable_ptr[p] = new Map<Address, CacheMsg>; m_readRequestTable_ptr[p] = new Map<Address, CacheMsg>;
@ -603,7 +606,8 @@ void Sequencer::hitCallback(const CacheMsg& request, DataBlock& data, GenericMac
(type == CacheRequestType_ATOMIC); (type == CacheRequestType_ATOMIC);
if (TSO && write) { if (TSO && write) {
m_chip_ptr->m_L1Cache_storeBuffer_vec[m_version]->callBack(line_address(request.getAddress()), data); m_chip_ptr->m_L1Cache_storeBuffer_vec[m_version]->callBack(line_address(request.getAddress()), data,
m_packetTable_ptr->lookup(request.getAddress()));
} else { } else {
// Copy the correct bytes out of the cache line into the subblock // Copy the correct bytes out of the cache line into the subblock
@ -616,7 +620,23 @@ void Sequencer::hitCallback(const CacheMsg& request, DataBlock& data, GenericMac
} }
// Call into the Driver and let it read and/or modify the sub-block // Call into the Driver and let it read and/or modify the sub-block
g_system_ptr->getDriver()->hitCallback(m_chip_ptr->getID()*RubyConfig::numberOfProcsPerChip()+m_version, subblock, type, threadID); Packet* pkt = m_packetTable_ptr->lookup(request.getAddress());
// update data if this is a store/atomic
/*
if (pkt->req->isCondSwap()) {
L1Cache_Entry entry = m_L1Cache_vec[m_version]->lookup(Address(pkt->req->physAddr()));
DataBlk datablk = entry->getDataBlk();
uint8_t *orig_data = datablk.getArray();
if ( datablk.equal(pkt->req->getExtraData()) )
datablk->setArray(pkt->getData());
pkt->setData(orig_data);
}
*/
g_system_ptr->getDriver()->hitCallback(pkt);
m_packetTable_ptr->remove(request.getAddress());
// If the request was a Store or Atomic, apply the changes in the SubBlock to the DataBlock // If the request was a Store or Atomic, apply the changes in the SubBlock to the DataBlock
// (This is only triggered for the non-TSO case) // (This is only triggered for the non-TSO case)
@ -632,6 +652,7 @@ void Sequencer::printDebug(){
g_system_ptr->getDriver()->printDebug(); g_system_ptr->getDriver()->printDebug();
} }
//dsm: breaks build, delayed
// Returns true if the sequencer already has a load or store outstanding // Returns true if the sequencer already has a load or store outstanding
bool bool
Sequencer::isReady(const Packet* pkt) const Sequencer::isReady(const Packet* pkt) const
@ -665,7 +686,7 @@ Sequencer::isReady(const Packet* pkt) const
Address(logical_addr), // Virtual Address Address(logical_addr), // Virtual Address
thread // SMT thread thread // SMT thread
); );
isReady(request); return isReady(request);
} }
bool bool
@ -701,26 +722,36 @@ Sequencer::isReady(const CacheMsg& request) const
return true; return true;
} }
// Called by Driver //dsm: breaks build, delayed
// Called by Driver (Simics or Tester).
void void
Sequencer::makeRequest(const Packet* pkt, void* data) Sequencer::makeRequest(Packet* pkt)
{ {
int cpu_number = pkt->req->contextId(); int cpu_number = pkt->req->contextId();
la_t logical_addr = pkt->req->getVaddr(); la_t logical_addr = pkt->req->getVaddr();
pa_t physical_addr = pkt->req->getPaddr(); pa_t physical_addr = pkt->req->getPaddr();
int request_size = pkt->getSize(); int request_size = pkt->getSize();
CacheRequestType type_of_request; CacheRequestType type_of_request;
PrefetchBit prefetch;
bool write = false;
if ( pkt->req->isInstFetch() ) { if ( pkt->req->isInstFetch() ) {
type_of_request = CacheRequestType_IFETCH; type_of_request = CacheRequestType_IFETCH;
} else if ( pkt->req->isLocked() || pkt->req->isSwap() ) { } else if ( pkt->req->isLocked() || pkt->req->isSwap() ) {
type_of_request = CacheRequestType_ATOMIC; type_of_request = CacheRequestType_ATOMIC;
write = true;
} else if ( pkt->isRead() ) { } else if ( pkt->isRead() ) {
type_of_request = CacheRequestType_LD; type_of_request = CacheRequestType_LD;
} else if ( pkt->isWrite() ) { } else if ( pkt->isWrite() ) {
type_of_request = CacheRequestType_ST; type_of_request = CacheRequestType_ST;
write = true;
} else { } else {
assert(false); assert(false);
} }
if (pkt->req->isPrefetch()) {
prefetch = PrefetchBit_Yes;
} else {
prefetch = PrefetchBit_No;
}
la_t virtual_pc = pkt->req->getPC(); la_t virtual_pc = pkt->req->getPC();
int isPriv = false; // TODO: get permission data int isPriv = false; // TODO: get permission data
int thread = pkt->req->threadId(); int thread = pkt->req->threadId();
@ -733,28 +764,21 @@ Sequencer::makeRequest(const Packet* pkt, void* data)
Address(virtual_pc), Address(virtual_pc),
access_mode, // User/supervisor mode access_mode, // User/supervisor mode
request_size, // Size in bytes of request request_size, // Size in bytes of request
PrefetchBit_No, // Not a prefetch prefetch,
0, // Version number 0, // Version number
Address(logical_addr), // Virtual Address Address(logical_addr), // Virtual Address
thread // SMT thread thread // SMT thread
); );
makeRequest(request);
}
void if ( TSO && write && !pkt->req->isPrefetch() ) {
Sequencer::makeRequest(const CacheMsg& request)
{
bool write = (request.getType() == CacheRequestType_ST) ||
(request.getType() == CacheRequestType_ATOMIC);
if (TSO && (request.getPrefetch() == PrefetchBit_No) && write) {
assert(m_chip_ptr->m_L1Cache_storeBuffer_vec[m_version]->isReady()); assert(m_chip_ptr->m_L1Cache_storeBuffer_vec[m_version]->isReady());
m_chip_ptr->m_L1Cache_storeBuffer_vec[m_version]->insertStore(request); m_chip_ptr->m_L1Cache_storeBuffer_vec[m_version]->insertStore(pkt, request);
return; return;
} }
bool hit = doRequest(request); m_packetTable_ptr->insert(Address( physical_addr ), pkt);
doRequest(request);
} }
bool Sequencer::doRequest(const CacheMsg& request) { bool Sequencer::doRequest(const CacheMsg& request) {

View file

@ -45,13 +45,13 @@
#include "mem/protocol/GenericMachineType.hh" #include "mem/protocol/GenericMachineType.hh"
#include "mem/protocol/PrefetchBit.hh" #include "mem/protocol/PrefetchBit.hh"
#include "mem/gems_common/Map.hh" #include "mem/gems_common/Map.hh"
#include "mem/packet.hh"
class DataBlock; class DataBlock;
class AbstractChip; class AbstractChip;
class CacheMsg; class CacheMsg;
class Address; class Address;
class MachineID; class MachineID;
class Packet;
class Sequencer : public Consumer { class Sequencer : public Consumer {
public: public:
@ -103,8 +103,7 @@ public:
void printDebug(); void printDebug();
// called by Tester or Simics // called by Tester or Simics
void makeRequest(const Packet* pkt, void* data); void makeRequest(Packet* pkt);
void makeRequest(const CacheMsg& request); // depricate this function
bool doRequest(const CacheMsg& request); bool doRequest(const CacheMsg& request);
void issueRequest(const CacheMsg& request); void issueRequest(const CacheMsg& request);
bool isReady(const Packet* pkt) const; bool isReady(const Packet* pkt) const;
@ -143,6 +142,9 @@ private:
// One request table per SMT thread // One request table per SMT thread
Map<Address, CacheMsg>** m_writeRequestTable_ptr; Map<Address, CacheMsg>** m_writeRequestTable_ptr;
Map<Address, CacheMsg>** m_readRequestTable_ptr; Map<Address, CacheMsg>** m_readRequestTable_ptr;
Map<Address, Packet*>* m_packetTable_ptr;
// Global outstanding request count, across all request tables // Global outstanding request count, across all request tables
int m_outstanding_count; int m_outstanding_count;
bool m_deadlock_check_scheduled; bool m_deadlock_check_scheduled;

View file

@ -44,6 +44,7 @@
#include "mem/ruby/system/Sequencer.hh" #include "mem/ruby/system/Sequencer.hh"
#include "mem/ruby/common/SubBlock.hh" #include "mem/ruby/common/SubBlock.hh"
#include "mem/ruby/profiler/Profiler.hh" #include "mem/ruby/profiler/Profiler.hh"
#include "mem/packet.hh"
// *** Begin Helper class *** // *** Begin Helper class ***
struct StoreBufferEntry { struct StoreBufferEntry {
@ -150,7 +151,8 @@ void StoreBuffer::printConfig(ostream& out)
// Handle an incoming store request, this method is responsible for // Handle an incoming store request, this method is responsible for
// calling hitCallback as needed // calling hitCallback as needed
void StoreBuffer::insertStore(const CacheMsg& request) void
StoreBuffer::insertStore(Packet* pkt, const CacheMsg& request)
{ {
Address addr = request.getAddress(); Address addr = request.getAddress();
CacheRequestType type = request.getType(); CacheRequestType type = request.getType();
@ -173,7 +175,7 @@ void StoreBuffer::insertStore(const CacheMsg& request)
// Perform the hit-callback for the store // Perform the hit-callback for the store
SubBlock subblock(addr, size); SubBlock subblock(addr, size);
if(type == CacheRequestType_ST) { if(type == CacheRequestType_ST) {
g_system_ptr->getDriver()->hitCallback(m_chip_ptr->getID(), subblock, type, threadID); g_system_ptr->getDriver()->hitCallback(pkt);
assert(subblock.getSize() != 0); assert(subblock.getSize() != 0);
} else { } else {
// wait to perform the hitCallback until later for Atomics // wait to perform the hitCallback until later for Atomics
@ -181,9 +183,9 @@ void StoreBuffer::insertStore(const CacheMsg& request)
// Perform possible pre-fetch // Perform possible pre-fetch
if(!isEmpty()) { if(!isEmpty()) {
CacheMsg new_request = request; Packet new_pkt(pkt);
new_request.getPrefetch() = PrefetchBit_Yes; pkt->req->setFlags(Request::PREFETCH);
m_chip_ptr->getSequencer(m_version)->makeRequest(new_request); m_chip_ptr->getSequencer(m_version)->makeRequest(&new_pkt);
} }
// Update the StoreCache // Update the StoreCache
@ -200,7 +202,7 @@ void StoreBuffer::insertStore(const CacheMsg& request)
processHeadOfQueue(); processHeadOfQueue();
} }
void StoreBuffer::callBack(const Address& addr, DataBlock& data) void StoreBuffer::callBack(const Address& addr, DataBlock& data, Packet* pkt)
{ {
DEBUG_MSG(STOREBUFFER_COMP, MedPrio, "callBack"); DEBUG_MSG(STOREBUFFER_COMP, MedPrio, "callBack");
DEBUG_EXPR(STOREBUFFER_COMP, MedPrio, g_eventQueue_ptr->getTime()); DEBUG_EXPR(STOREBUFFER_COMP, MedPrio, g_eventQueue_ptr->getTime());
@ -220,7 +222,7 @@ void StoreBuffer::callBack(const Address& addr, DataBlock& data)
} else { } else {
// We waited to perform the hitCallback until now for Atomics // We waited to perform the hitCallback until now for Atomics
peek().m_subblock.mergeFrom(data); // copy the correct bytes from DataBlock into the SubBlock for the Load part of the atomic Load/Store peek().m_subblock.mergeFrom(data); // copy the correct bytes from DataBlock into the SubBlock for the Load part of the atomic Load/Store
g_system_ptr->getDriver()->hitCallback(m_chip_ptr->getID(), peek().m_subblock, type, threadID); g_system_ptr->getDriver()->hitCallback(pkt);
m_seen_atomic = false; m_seen_atomic = false;
/// FIXME - record the time spent in the store buffer - split out ST vs ATOMIC /// FIXME - record the time spent in the store buffer - split out ST vs ATOMIC

View file

@ -49,6 +49,7 @@ class DataBlock;
class SubBlock; class SubBlock;
class StoreBufferEntry; class StoreBufferEntry;
class AbstractChip; class AbstractChip;
class Packet;
template <class TYPE> class Vector; template <class TYPE> class Vector;
@ -62,8 +63,8 @@ public:
// Public Methods // Public Methods
void wakeup(); // Used only for deadlock detection void wakeup(); // Used only for deadlock detection
void callBack(const Address& addr, DataBlock& data); void callBack(const Address& addr, DataBlock& data, Packet* pkt);
void insertStore(const CacheMsg& request); void insertStore(Packet* pkt, const CacheMsg& request);
void updateSubBlock(SubBlock& sub_block) const { m_store_cache.update(sub_block); } void updateSubBlock(SubBlock& sub_block) const { m_store_cache.update(sub_block); }
bool trySubBlock(const SubBlock& sub_block) const { assert(isReady()); return m_store_cache.check(sub_block); } bool trySubBlock(const SubBlock& sub_block) const { assert(isReady()); return m_store_cache.check(sub_block); }
void print(ostream& out) const; void print(ostream& out) const;

View file

@ -46,12 +46,48 @@
#include "mem/protocol/Chip.hh" #include "mem/protocol/Chip.hh"
//#include "mem/ruby/recorder/Tracer.hh" //#include "mem/ruby/recorder/Tracer.hh"
#include "mem/protocol/Protocol.hh" #include "mem/protocol/Protocol.hh"
//#include "XactIsolationChecker.hh" // gem5:Arka for decomissioning of log_tm
//#include "XactCommitArbiter.hh"
//#include "XactVisualizer.hh"
#include "mem/ruby/interfaces/M5Driver.hh"
RubySystem::RubySystem() RubySystem::RubySystem()
{
init();
m_preinitialized_driver = false;
createDriver();
/* gem5:Binkert for decomissiong of tracer
m_tracer_ptr = new Tracer;
*/
/* gem5:Arka for decomissiong of log_tm
if (XACT_MEMORY) {
m_xact_isolation_checker = new XactIsolationChecker;
m_xact_commit_arbiter = new XactCommitArbiter;
m_xact_visualizer = new XactVisualizer;
}
*/
}
RubySystem::RubySystem(Driver* _driver)
{
init();
m_preinitialized_driver = true;
m_driver_ptr = _driver;
}
RubySystem::~RubySystem()
{
for (int i = 0; i < m_chip_vector.size(); i++) {
delete m_chip_vector[i];
}
if (!m_preinitialized_driver)
delete m_driver_ptr;
delete m_network_ptr;
delete m_profiler_ptr;
/* gem5:Binkert for decomissiong of tracer
delete m_tracer_ptr;
*/
}
void RubySystem::init()
{ {
DEBUG_MSG(SYSTEM_COMP, MedPrio,"initializing"); DEBUG_MSG(SYSTEM_COMP, MedPrio,"initializing");
@ -101,44 +137,19 @@ RubySystem::RubySystem()
} }
} }
#endif #endif
DEBUG_MSG(SYSTEM_COMP, MedPrio,"finished initializing");
DEBUG_NEWLINE(SYSTEM_COMP, MedPrio);
}
void RubySystem::createDriver()
{
if (g_SYNTHETIC_DRIVER && !g_DETERMINISTIC_DRIVER) { if (g_SYNTHETIC_DRIVER && !g_DETERMINISTIC_DRIVER) {
cerr << "Creating Synthetic Driver" << endl; cerr << "Creating Synthetic Driver" << endl;
m_driver_ptr = new SyntheticDriver(this); m_driver_ptr = new SyntheticDriver(this);
} else if (!g_SYNTHETIC_DRIVER && g_DETERMINISTIC_DRIVER) { } else if (!g_SYNTHETIC_DRIVER && g_DETERMINISTIC_DRIVER) {
cerr << "Creating Deterministic Driver" << endl; cerr << "Creating Deterministic Driver" << endl;
m_driver_ptr = new DeterministicDriver(this); m_driver_ptr = new DeterministicDriver(this);
} else {
cerr << "Creating M5 Driver" << endl;
m_driver_ptr = new M5Driver(this);
} }
/* gem5:Binkert for decomissiong of tracer
m_tracer_ptr = new Tracer;
*/
/* gem5:Arka for decomissiong of log_tm
if (XACT_MEMORY) {
m_xact_isolation_checker = new XactIsolationChecker;
m_xact_commit_arbiter = new XactCommitArbiter;
m_xact_visualizer = new XactVisualizer;
}
*/
DEBUG_MSG(SYSTEM_COMP, MedPrio,"finished initializing");
DEBUG_NEWLINE(SYSTEM_COMP, MedPrio);
}
RubySystem::~RubySystem()
{
for (int i = 0; i < m_chip_vector.size(); i++) {
delete m_chip_vector[i];
}
delete m_driver_ptr;
delete m_network_ptr;
delete m_profiler_ptr;
/* gem5:Binkert for decomissiong of tracer
delete m_tracer_ptr;
*/
} }
void RubySystem::printConfig(ostream& out) const void RubySystem::printConfig(ostream& out) const

View file

@ -63,6 +63,7 @@ class RubySystem {
public: public:
// Constructors // Constructors
RubySystem(); RubySystem();
RubySystem(Driver* _driver); // used when driver is already instantiated (e.g. M5's RubyMem)
// Destructor // Destructor
~RubySystem(); ~RubySystem();
@ -98,6 +99,8 @@ public:
private: private:
// Private Methods // Private Methods
void init();
void createDriver();
// Private copy constructor and assignment operator // Private copy constructor and assignment operator
RubySystem(const RubySystem& obj); RubySystem(const RubySystem& obj);
@ -107,6 +110,7 @@ private:
Network* m_network_ptr; Network* m_network_ptr;
Vector<AbstractChip*> m_chip_vector; Vector<AbstractChip*> m_chip_vector;
Profiler* m_profiler_ptr; Profiler* m_profiler_ptr;
bool m_preinitialized_driver;
Driver* m_driver_ptr; Driver* m_driver_ptr;
Tracer* m_tracer_ptr; Tracer* m_tracer_ptr;
XactIsolationChecker *m_xact_isolation_checker; XactIsolationChecker *m_xact_isolation_checker;

View file

@ -37,6 +37,7 @@
#include "mem/ruby/system/System.hh" #include "mem/ruby/system/System.hh"
#include "mem/ruby/common/SubBlock.hh" #include "mem/ruby/common/SubBlock.hh"
#include "mem/protocol/Chip.hh" #include "mem/protocol/Chip.hh"
#include "mem/packet.hh"
Check::Check(const Address& address, const Address& pc) Check::Check(const Address& address, const Address& pc)
{ {
@ -84,10 +85,29 @@ void Check::initiatePrefetch(Sequencer* targetSequencer_ptr)
} else { } else {
type = CacheRequestType_ST; type = CacheRequestType_ST;
} }
Addr data_addr = m_address.getAddress();
Addr pc_addr = m_pc.getAddress();
Request request(0, data_addr, 0, Flags<unsigned int>(Request::PREFETCH), pc_addr, 0, 0);
MemCmd::Command command;
if (type == CacheRequestType_IFETCH) {
command = MemCmd::ReadReq;
request.setFlags(Request::INST_FETCH);
} else if (type == CacheRequestType_LD || type == CacheRequestType_IFETCH) {
command = MemCmd::ReadReq;
} else if (type == CacheRequestType_ST) {
command = MemCmd::WriteReq;
} else if (type == CacheRequestType_ATOMIC) {
command = MemCmd::SwapReq; // TODO -- differentiate between atomic types
} else {
assert(false);
}
Packet pkt(&request, command, 0); // TODO -- make dest a real NodeID
assert(targetSequencer_ptr != NULL); assert(targetSequencer_ptr != NULL);
CacheMsg request(m_address, m_address, type, m_pc, m_access_mode, 0, PrefetchBit_Yes, 0, Address(0), 0 /* only 1 SMT thread */); if (targetSequencer_ptr->isReady(&pkt)) {
if (targetSequencer_ptr->isReady(request)) { targetSequencer_ptr->makeRequest(&pkt);
targetSequencer_ptr->makeRequest(request);
} }
} }
@ -109,15 +129,34 @@ void Check::initiateAction()
type = CacheRequestType_ATOMIC; type = CacheRequestType_ATOMIC;
} }
CacheMsg request(Address(m_address.getAddress()+m_store_count), Address(m_address.getAddress()+m_store_count), type, m_pc, m_access_mode, 1, PrefetchBit_No, 0, Address(0), 0 /* only 1 SMT thread */); Addr data_addr = m_address.getAddress()+m_store_count;
Addr pc_addr = m_pc.getAddress();
Request request(0, data_addr, 1, Flags<unsigned int>(), pc_addr, 0, 0);
MemCmd::Command command;
if (type == CacheRequestType_IFETCH) {
command = MemCmd::ReadReq;
request.setFlags(Request::INST_FETCH);
} else if (type == CacheRequestType_LD || type == CacheRequestType_IFETCH) {
command = MemCmd::ReadReq;
} else if (type == CacheRequestType_ST) {
command = MemCmd::WriteReq;
} else if (type == CacheRequestType_ATOMIC) {
command = MemCmd::SwapReq; // TODO -- differentiate between atomic types
} else {
assert(false);
}
Packet pkt(&request, command, 0); // TODO -- make dest a real NodeID
Sequencer* sequencer_ptr = initiatingSequencer(); Sequencer* sequencer_ptr = initiatingSequencer();
if (sequencer_ptr->isReady(request) == false) { if (sequencer_ptr->isReady(&pkt) == false) {
DEBUG_MSG(TESTER_COMP, MedPrio, "failed to initiate action - sequencer not ready\n"); DEBUG_MSG(TESTER_COMP, MedPrio, "failed to initiate action - sequencer not ready\n");
} else { } else {
DEBUG_MSG(TESTER_COMP, MedPrio, "initiating action - successful\n"); DEBUG_MSG(TESTER_COMP, MedPrio, "initiating action - successful\n");
DEBUG_EXPR(TESTER_COMP, MedPrio, m_status); DEBUG_EXPR(TESTER_COMP, MedPrio, m_status);
m_status = TesterStatus_Action_Pending; m_status = TesterStatus_Action_Pending;
sequencer_ptr->makeRequest(request);
sequencer_ptr->makeRequest(&pkt);
} }
DEBUG_EXPR(TESTER_COMP, MedPrio, m_status); DEBUG_EXPR(TESTER_COMP, MedPrio, m_status);
} }
@ -132,15 +171,35 @@ void Check::initiateCheck()
type = CacheRequestType_IFETCH; type = CacheRequestType_IFETCH;
} }
CacheMsg request(m_address, m_address, type, m_pc, m_access_mode, CHECK_SIZE, PrefetchBit_No, 0, Address(0), 0 /* only 1 SMT thread */);
Addr data_addr = m_address.getAddress()+m_store_count;
Addr pc_addr = m_pc.getAddress();
Request request(0, data_addr, CHECK_SIZE, Flags<unsigned int>(), pc_addr, 0, 0);
MemCmd::Command command;
if (type == CacheRequestType_IFETCH) {
command = MemCmd::ReadReq;
request.setFlags(Request::INST_FETCH);
} else if (type == CacheRequestType_LD || type == CacheRequestType_IFETCH) {
command = MemCmd::ReadReq;
} else if (type == CacheRequestType_ST) {
command = MemCmd::WriteReq;
} else if (type == CacheRequestType_ATOMIC) {
command = MemCmd::SwapReq; // TODO -- differentiate between atomic types
} else {
assert(false);
}
Packet pkt(&request, command, 0); // TODO -- make dest a real NodeID
Sequencer* sequencer_ptr = initiatingSequencer(); Sequencer* sequencer_ptr = initiatingSequencer();
if (sequencer_ptr->isReady(request) == false) { if (sequencer_ptr->isReady(&pkt) == false) {
DEBUG_MSG(TESTER_COMP, MedPrio, "failed to initiate check - sequencer not ready\n"); DEBUG_MSG(TESTER_COMP, MedPrio, "failed to initiate check - sequencer not ready\n");
} else { } else {
DEBUG_MSG(TESTER_COMP, MedPrio, "initiating check - successful\n"); DEBUG_MSG(TESTER_COMP, MedPrio, "initiating check - successful\n");
DEBUG_MSG(TESTER_COMP, MedPrio, m_status); DEBUG_MSG(TESTER_COMP, MedPrio, m_status);
m_status = TesterStatus_Check_Pending; m_status = TesterStatus_Check_Pending;
sequencer_ptr->makeRequest(request);
sequencer_ptr->makeRequest(&pkt);
} }
DEBUG_MSG(TESTER_COMP, MedPrio, m_status); DEBUG_MSG(TESTER_COMP, MedPrio, m_status);
} }

View file

@ -44,6 +44,7 @@
#include "mem/ruby/common/SubBlock.hh" #include "mem/ruby/common/SubBlock.hh"
#include "mem/ruby/tester/DeterministicDriver.hh" #include "mem/ruby/tester/DeterministicDriver.hh"
#include "mem/protocol/Chip.hh" #include "mem/protocol/Chip.hh"
#include "mem/packet.hh"
DetermGETXGenerator::DetermGETXGenerator(NodeID node, DeterministicDriver& driver) : DetermGETXGenerator::DetermGETXGenerator(NodeID node, DeterministicDriver& driver) :
m_driver(driver) m_driver(driver)
@ -137,7 +138,15 @@ void DetermGETXGenerator::pickAddress()
void DetermGETXGenerator::initiateStore() void DetermGETXGenerator::initiateStore()
{ {
DEBUG_MSG(TESTER_COMP, MedPrio, "initiating Store"); DEBUG_MSG(TESTER_COMP, MedPrio, "initiating Store");
sequencer()->makeRequest(CacheMsg(m_address, m_address, CacheRequestType_ST, Address(3), AccessModeType_UserMode, 1, PrefetchBit_No, 0, Address(0), 0 /* only 1 SMT thread */));
Addr data_addr = m_address.getAddress();
Request request(0, data_addr, 1, Flags<unsigned int>(), 3, 0, 0);
MemCmd::Command command;
command = MemCmd::WriteReq;
Packet pkt(&request, command, 0); // TODO -- make dest a real NodeID
sequencer()->makeRequest(&pkt);
} }
Sequencer* DetermGETXGenerator::sequencer() const Sequencer* DetermGETXGenerator::sequencer() const

View file

@ -179,13 +179,30 @@ void DetermInvGenerator::pickLoadAddress()
void DetermInvGenerator::initiateLoad() void DetermInvGenerator::initiateLoad()
{ {
DEBUG_MSG(TESTER_COMP, MedPrio, "initiating Load"); DEBUG_MSG(TESTER_COMP, MedPrio, "initiating Load");
sequencer()->makeRequest(CacheMsg(m_address, m_address, CacheRequestType_LD, Address(1), AccessModeType_UserMode, 1, PrefetchBit_No, 0, Address(0), 0 /* only 1 SMT thread */));
Addr data_addr = m_address.getAddress();
Request request(0, data_addr, 1, Flags<unsigned int>(), 1, 0, 0);
MemCmd::Command command;
command = MemCmd::ReadReq;
Packet pkt(&request, command, 0); // TODO -- make dest a real NodeID
sequencer()->makeRequest(&pkt);
} }
void DetermInvGenerator::initiateStore() void DetermInvGenerator::initiateStore()
{ {
DEBUG_MSG(TESTER_COMP, MedPrio, "initiating Store"); DEBUG_MSG(TESTER_COMP, MedPrio, "initiating Store");
sequencer()->makeRequest(CacheMsg(m_address, m_address, CacheRequestType_ST, Address(3), AccessModeType_UserMode, 1, PrefetchBit_No, 0, Address(0), 0 /* only 1 SMT thread */));
Addr data_addr = m_address.getAddress();
Request request(0, data_addr, 1, Flags<unsigned int>(), 3, 0, 0);
MemCmd::Command command;
command = MemCmd::WriteReq;
Packet pkt(&request, command, 0); // TODO -- make dest a real NodeID
sequencer()->makeRequest(&pkt);
} }
Sequencer* DetermInvGenerator::sequencer() const Sequencer* DetermInvGenerator::sequencer() const

View file

@ -135,7 +135,16 @@ void DetermSeriesGETSGenerator::pickAddress()
void DetermSeriesGETSGenerator::initiateLoad() void DetermSeriesGETSGenerator::initiateLoad()
{ {
DEBUG_MSG(TESTER_COMP, MedPrio, "initiating Load"); DEBUG_MSG(TESTER_COMP, MedPrio, "initiating Load");
sequencer()->makeRequest(CacheMsg(m_address, m_address, CacheRequestType_IFETCH, Address(3), AccessModeType_UserMode, 1, PrefetchBit_No, 0, Address(0), 0 /* only 1 SMT thread */));
Addr data_addr = m_address.getAddress();
Request request(0, data_addr, 1, Flags<unsigned int>(), 3, 0, 0);
MemCmd::Command command;
command = MemCmd::ReadReq;
request.setFlags(Request::INST_FETCH);
Packet pkt(&request, command, 0); // TODO -- make dest a real NodeID
sequencer()->makeRequest(&pkt);
} }
Sequencer* DetermSeriesGETSGenerator::sequencer() const Sequencer* DetermSeriesGETSGenerator::sequencer() const

View file

@ -42,6 +42,7 @@
#include "mem/ruby/tester/DetermSeriesGETSGenerator.hh" #include "mem/ruby/tester/DetermSeriesGETSGenerator.hh"
#include "mem/ruby/common/SubBlock.hh" #include "mem/ruby/common/SubBlock.hh"
#include "mem/protocol/Chip.hh" #include "mem/protocol/Chip.hh"
#include "mem/packet.hh"
DeterministicDriver::DeterministicDriver(RubySystem* sys_ptr) DeterministicDriver::DeterministicDriver(RubySystem* sys_ptr)
{ {
@ -99,13 +100,17 @@ DeterministicDriver::~DeterministicDriver()
} }
} }
void DeterministicDriver::hitCallback(NodeID proc, SubBlock& data, CacheRequestType type, int thread) void
DeterministicDriver::hitCallback(Packet * pkt)
{ {
DEBUG_EXPR(TESTER_COMP, MedPrio, data); NodeID proc = pkt->req->contextId();
SubBlock data(Address(pkt->getAddr()), pkt->req->getSize());
if (pkt->hasData()) {
for (int i = 0; i < pkt->req->getSize(); i++) {
data.setByte(i, *(pkt->getPtr<uint8>()+i));
}
}
m_generator_vector[proc]->performCallback(proc, data); m_generator_vector[proc]->performCallback(proc, data);
// Mark that we made progress
m_last_progress_vector[proc] = g_eventQueue_ptr->getTime(); m_last_progress_vector[proc] = g_eventQueue_ptr->getTime();
} }

View file

@ -44,6 +44,7 @@
class RubySystem; class RubySystem;
class SpecifiedGenerator; class SpecifiedGenerator;
class Packet;
class DeterministicDriver : public Driver, public Consumer { class DeterministicDriver : public Driver, public Consumer {
public: public:
@ -69,7 +70,7 @@ public:
void recordLoadLatency(Time time); void recordLoadLatency(Time time);
void recordStoreLatency(Time time); void recordStoreLatency(Time time);
void hitCallback(NodeID proc, SubBlock& data, CacheRequestType type, int thread); void hitCallback(Packet* pkt);
void wakeup(); void wakeup();
void printStats(ostream& out) const; void printStats(ostream& out) const;
void clearStats() {} void clearStats() {}

View file

@ -169,19 +169,43 @@ void RequestGenerator::pickAddress()
void RequestGenerator::initiateTest() void RequestGenerator::initiateTest()
{ {
DEBUG_MSG(TESTER_COMP, MedPrio, "initiating Test"); DEBUG_MSG(TESTER_COMP, MedPrio, "initiating Test");
sequencer()->makeRequest(CacheMsg(m_address, m_address, CacheRequestType_LD, Address(1), AccessModeType_UserMode, 1, PrefetchBit_No, 0, Address(0), 0 /* only 1 SMT thread */));
Addr data_addr = m_address.getAddress();
Request request(0, data_addr, 1, Flags<unsigned int>(), 1, 0, 0);
MemCmd::Command command;
command = MemCmd::ReadReq;
Packet pkt(&request, command, 0); // TODO -- make dest a real NodeID
sequencer()->makeRequest(&pkt);
} }
void RequestGenerator::initiateSwap() void RequestGenerator::initiateSwap()
{ {
DEBUG_MSG(TESTER_COMP, MedPrio, "initiating Swap"); DEBUG_MSG(TESTER_COMP, MedPrio, "initiating Swap");
sequencer()->makeRequest(CacheMsg(m_address, m_address, CacheRequestType_ATOMIC, Address(2), AccessModeType_UserMode, 1, PrefetchBit_No, 0, Address(0), 0 /* only 1 SMT thread */));
Addr data_addr = m_address.getAddress();
Request request(0, data_addr, 1, Flags<unsigned int>(), 2, 0, 0);
MemCmd::Command command;
command = MemCmd::SwapReq;
Packet pkt(&request, command, 0); // TODO -- make dest a real NodeID
sequencer()->makeRequest(&pkt);
} }
void RequestGenerator::initiateRelease() void RequestGenerator::initiateRelease()
{ {
DEBUG_MSG(TESTER_COMP, MedPrio, "initiating Release"); DEBUG_MSG(TESTER_COMP, MedPrio, "initiating Release");
sequencer()->makeRequest(CacheMsg(m_address, m_address, CacheRequestType_ST, Address(3), AccessModeType_UserMode, 1, PrefetchBit_No, 0, Address(0), 0 /* only 1 SMT thread */));
Addr data_addr = m_address.getAddress();
Request request(0, data_addr, 1, Flags<unsigned int>(), 3, 0, 0);
MemCmd::Command command;
command = MemCmd::WriteReq;
Packet pkt(&request, command, 0); // TODO -- make dest a real NodeID
sequencer()->makeRequest(&pkt);
} }
Sequencer* RequestGenerator::sequencer() const Sequencer* RequestGenerator::sequencer() const

View file

@ -75,18 +75,17 @@ SyntheticDriver::~SyntheticDriver()
} }
} }
void SyntheticDriver::hitCallback(NodeID proc, SubBlock& data, CacheRequestType type, int thread) void
SyntheticDriver::hitCallback(Packet * pkt)
{ {
DEBUG_EXPR(TESTER_COMP, MedPrio, data); NodeID proc = pkt->req->contextId();
//cout << " " << proc << " in S.D. hitCallback" << endl; SubBlock data(Address(pkt->getAddr()), pkt->req->getSize());
if(XACT_MEMORY){ if (pkt->hasData()) {
//XactRequestGenerator* reqGen = static_cast<XactRequestGenerator*>(m_request_generator_vector[proc]); for (int i = 0; i < pkt->req->getSize(); i++) {
//reqGen->performCallback(proc, data); data.setByte(i, *(pkt->getPtr<uint8>()+i));
} else {
m_request_generator_vector[proc]->performCallback(proc, data);
} }
}
// Mark that we made progress m_request_generator_vector[proc]->performCallback(proc, data);
m_last_progress_vector[proc] = g_eventQueue_ptr->getTime(); m_last_progress_vector[proc] = g_eventQueue_ptr->getTime();
} }

View file

@ -60,7 +60,7 @@ public:
void recordSwapLatency(Time time); void recordSwapLatency(Time time);
void recordReleaseLatency(Time time); void recordReleaseLatency(Time time);
void hitCallback(NodeID proc, SubBlock& data, CacheRequestType type, int thread); void hitCallback(Packet* pkt);
void conflictCallback(NodeID proc, SubBlock& data, CacheRequestType type, int thread) {assert(0);} void conflictCallback(NodeID proc, SubBlock& data, CacheRequestType type, int thread) {assert(0);}
void abortCallback(NodeID proc, SubBlock& data, CacheRequestType type, int thread); void abortCallback(NodeID proc, SubBlock& data, CacheRequestType type, int thread);
void wakeup(); void wakeup();

View file

@ -35,7 +35,7 @@
#include "mem/ruby/tester/main.hh" #include "mem/ruby/tester/main.hh"
#include "mem/ruby/eventqueue/RubyEventQueue.hh" #include "mem/ruby/eventqueue/RubyEventQueue.hh"
#include "mem/ruby/config/RubyConfig.hh" #include "mem/ruby/config/RubyConfig.hh"
#include "mem/ruby/tester/test_framework.hh" //#include "mem/ruby/tester/test_framework.hh"
// ******************* // *******************
// *** tester main *** // *** tester main ***
@ -43,5 +43,6 @@
int main(int argc, char *argv[]) int main(int argc, char *argv[])
{ {
tester_main(argc, argv); //dsm: PRUNED
//tester_main(argc, argv);
} }

256
src/mem/rubymem.cc Normal file
View file

@ -0,0 +1,256 @@
/*
* Copyright (c) 2001-2005 The Regents of The University of Michigan
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met: redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer;
* redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution;
* neither the name of the copyright holders nor the names of its
* contributors may be used to endorse or promote products derived from
* this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* Authors: Daniel Sanchez
*/
#include "arch/isa_traits.hh"
#include "mem/rubymem.hh"
#include "sim/eventq.hh"
#include "sim/host.hh"
#include "base/output.hh"
// Ruby includes
#include "mem/ruby/system/System.hh"
#include "mem/ruby/system/Sequencer.hh"
#include "mem/ruby/init.hh"
#include "mem/ruby/common/Debug.hh"
#include "sim/sim_exit.hh"
#include <iostream>
#include <fstream>
using namespace std;
using namespace TheISA;
RubyMemory::RubyMemory(const Params *p)
: PhysicalMemory(p)
{
config_file = p->config_file;
config_options = p->config_options;
stats_file = p->stats_file;
num_cpus = p->num_cpus;
ruby_clock = p->clock;
ruby_phase = p->phase;
debug = p->debug;
debug_file = p->debug_file;
}
void
RubyMemory::init()
{
init_variables();
g_NUM_PROCESSORS = num_cpus;
init_simulator(this);
if (debug) {
g_debug_ptr->setVerbosityString("high");
g_debug_ptr->setDebugTime(1);
if (debug_file != "") {
g_debug_ptr->setDebugOutputFile("ruby.debug");
}
}
//You may want to set some other options...
//g_debug_ptr->setVerbosityString("med");
//g_debug_ptr->setFilterString("lsNqST");
//g_debug_ptr->setFilterString("lsNST");
//g_debug_ptr->setDebugTime(1);
//g_debug_ptr->setDebugOutputFile("ruby.debug");
g_system_ptr->clearStats();
if (ports.size() == 0) {
fatal("RubyMemory object %s is unconnected!", name());
}
for (PortIterator pi = ports.begin(); pi != ports.end(); ++pi) {
if (*pi)
(*pi)->sendStatusChange(Port::RangeChange);
}
//Print stats at exit
RubyExitCallback* rc = new RubyExitCallback(this);
registerExitCallback(rc);
//Sched RubyEvent, automatically reschedules to advance ruby cycles
rubyTickEvent = new RubyEvent(this);
schedule(rubyTickEvent, curTick + ruby_clock + ruby_phase);
}
//called by rubyTickEvent
void RubyMemory::tick() {
g_eventQueue_ptr->triggerEvents(g_eventQueue_ptr->getTime() + 1);
schedule(rubyTickEvent, curTick + ruby_clock); //dsm: clock_phase was added here. This is wrong, the phase is only added on the first tick
}
RubyMemory::~RubyMemory() {
delete g_system_ptr;
}
void
RubyMemory::hitCallback(Packet* pkt)
{
RubyMemoryPort* port = m_packet_to_port_map[pkt];
assert(port != NULL);
m_packet_to_port_map.erase(pkt);
DPRINTF(MemoryAccess, "Hit callback\n");
bool needsResponse = pkt->needsResponse();
doAtomicAccess(pkt);
// turn packet around to go back to requester if response expected
if (needsResponse) {
// recvAtomic() should already have turned packet into
// atomic response
assert(pkt->isResponse());
DPRINTF(MemoryAccess, "Sending packet back over port\n");
port->sendTiming(pkt);
} else {
delete pkt;
}
DPRINTF(MemoryAccess, "Hit callback done!\n");
}
Port *
RubyMemory::getPort(const std::string &if_name, int idx)
{
// Accept request for "functional" port for backwards compatibility
// with places where this function is called from C++. I'd prefer
// to move all these into Python someday.
if (if_name == "functional") {
return new RubyMemoryPort(csprintf("%s-functional", name()), this);
}
if (if_name != "port") {
panic("RubyMemory::getPort: unknown port %s requested", if_name);
}
if (idx >= ports.size()) {
ports.resize(idx+1);
}
if (ports[idx] != NULL) {
panic("RubyMemory::getPort: port %d already assigned", idx);
}
RubyMemoryPort *port =
new RubyMemoryPort(csprintf("%s-port%d", name(), idx), this);
ports[idx] = port;
return port;
}
RubyMemory::RubyMemoryPort::RubyMemoryPort(const std::string &_name,
RubyMemory *_memory)
: PhysicalMemory::MemoryPort::MemoryPort(_name, _memory)
{
ruby_mem = _memory;
}
bool
RubyMemory::RubyMemoryPort::recvTiming(PacketPtr pkt)
{
DPRINTF(MemoryAccess, "Timing access caught\n");
//dsm: based on SimpleTimingPort::recvTiming(pkt);
// If the device is only a slave, it should only be sending
// responses, which should never get nacked. There used to be
// code to hanldle nacks here, but I'm pretty sure it didn't work
// correctly with the drain code, so that would need to be fixed
// if we ever added it back.
assert(pkt->isRequest());
if (pkt->memInhibitAsserted()) {
warn("memInhibitAsserted???");
// snooper will supply based on copy of packet
// still target's responsibility to delete packet
delete pkt;
return true;
}
ruby_mem->m_packet_to_port_map[pkt] = this;
Sequencer* sequencer = g_system_ptr->getSequencer(pkt->req->contextId());
if ( ! sequencer->isReady(pkt) ) {
DPRINTF(MemoryAccess, "Sequencer isn't ready yet!!\n");
return false;
}
DPRINTF(MemoryAccess, "Issuing makeRequest\n");
sequencer->makeRequest(pkt);
return true;
}
void
RubyMemory::RubyMemoryPort::sendTiming(PacketPtr pkt)
{
schedSendTiming(pkt, curTick + 1); //minimum latency, must be > 0
}
void RubyMemory::printConfigStats()
{
std::ostream *os = simout.create(stats_file);
g_system_ptr->printConfig(*os);
*os << endl;
g_system_ptr->printStats(*os);
}
//Right now these functions seem to be called by RubySystem. If they do calls
// to RubySystem perform it intended actions, you'll get into an inf loop
//FIXME what's the purpose of these here?
void RubyMemory::printStats(std::ostream & out) const {
//g_system_ptr->printConfig(out);
}
void RubyMemory::clearStats() {
//g_system_ptr->clearStats();
}
void RubyMemory::printConfig(std::ostream & out) const {
//g_system_ptr->printConfig(out);
}
//Python-interface code
RubyMemory *
RubyMemoryParams::create()
{
return new RubyMemory(this);
}

129
src/mem/rubymem.hh Normal file
View file

@ -0,0 +1,129 @@
/*
* Copyright (c) 2001-2005 The Regents of The University of Michigan
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met: redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer;
* redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution;
* neither the name of the copyright holders nor the names of its
* contributors may be used to endorse or promote products derived from
* this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* Authors: Daniel Sanchez
*/
#ifndef __RUBY_MEMORY_HH__
#define __RUBY_MEMORY_HH__
#include <map>
#include "mem/physical.hh"
#include "params/RubyMemory.hh"
#include "base/callback.hh"
#include "mem/ruby/common/Driver.hh"
class RubyMemory : public PhysicalMemory, public Driver
{
class RubyMemoryPort : public MemoryPort
{
RubyMemory* ruby_mem;
public:
RubyMemoryPort(const std::string &_name, RubyMemory *_memory);
void sendTiming(PacketPtr pkt);
protected:
virtual bool recvTiming(PacketPtr pkt);
};
class RubyEvent : public Event
{
RubyMemory *ruby_ptr;
public:
RubyEvent(RubyMemory *p)
: Event(), ruby_ptr(p) {}
virtual void process() { ruby_ptr->tick(); }
virtual const char *description() const { return "ruby tick"; }
};
private:
// prevent copying of a RubyMemory object
RubyMemory(const RubyMemory &specmem);
const RubyMemory &operator=(const RubyMemory &specmem);
RubyEvent* rubyTickEvent;
public:
typedef RubyMemoryParams Params;
RubyMemory(const Params *p);
virtual ~RubyMemory();
public:
virtual Port *getPort(const std::string &if_name, int idx = -1);
void virtual init();
//Ruby-related specifics
void printConfigStats(); //dsm: Maybe this function should disappear once the configuration options change & M5 determines the stats file to use
void hitCallback(Packet* pkt); // called by the Ruby sequencer
void printStats(std::ostream & out) const;
void clearStats();
void printConfig(std::ostream & out) const;
void tick();
private:
//Parameters passed
std::string config_file, config_options, stats_file, debug_file;
bool debug;
int num_cpus;
Tick ruby_clock, ruby_phase;
std::map<Packet*, RubyMemoryPort*> m_packet_to_port_map;
};
class RubyExitCallback : public Callback
{
private:
RubyMemory* ruby;
public:
/**
* virtualize the destructor to make sure that the correct one
* gets called.
*/
virtual ~RubyExitCallback() {};
RubyExitCallback(RubyMemory* rm) {ruby=rm;};
/**
* virtual process function that is invoked when the callback
* queue is executed.
*/
virtual void process() {ruby->printConfigStats(); /*delete ruby; was doing double delete...*/};
};
#endif //__RUBY_MEMORY_HH__