gem5/src/mem/ruby/system/Sequencer.cc
2009-07-21 19:42:09 -05:00

494 lines
17 KiB
C++

/*
* 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.
*/
#include "mem/ruby/common/Global.hh"
#include "mem/ruby/system/Sequencer.hh"
#include "mem/ruby/system/System.hh"
#include "mem/protocol/Protocol.hh"
#include "mem/ruby/profiler/Profiler.hh"
#include "mem/ruby/system/CacheMemory.hh"
#include "mem/protocol/CacheMsg.hh"
#include "mem/ruby/recorder/Tracer.hh"
#include "mem/ruby/common/SubBlock.hh"
#include "mem/protocol/Protocol.hh"
#include "mem/gems_common/Map.hh"
#include "mem/ruby/buffers/MessageBuffer.hh"
#include "mem/ruby/slicc_interface/AbstractController.hh"
//Sequencer::Sequencer(int core_id, MessageBuffer* mandatory_q)
#define LLSC_FAIL -2
Sequencer::Sequencer(const string & name)
:RubyPort(name)
{
}
void Sequencer::init(const vector<string> & argv)
{
m_deadlock_check_scheduled = false;
m_outstanding_count = 0;
m_max_outstanding_requests = 0;
m_deadlock_threshold = 0;
m_version = -1;
m_instCache_ptr = NULL;
m_dataCache_ptr = NULL;
m_controller = NULL;
for (size_t i=0; i<argv.size(); i+=2) {
if ( argv[i] == "controller") {
m_controller = RubySystem::getController(argv[i+1]); // args[i] = "L1Cache"
m_mandatory_q_ptr = m_controller->getMandatoryQueue();
} else if ( argv[i] == "icache")
m_instCache_ptr = RubySystem::getCache(argv[i+1]);
else if ( argv[i] == "dcache")
m_dataCache_ptr = RubySystem::getCache(argv[i+1]);
else if ( argv[i] == "version")
m_version = atoi(argv[i+1].c_str());
else if ( argv[i] == "max_outstanding_requests")
m_max_outstanding_requests = atoi(argv[i+1].c_str());
else if ( argv[i] == "deadlock_threshold")
m_deadlock_threshold = atoi(argv[i+1].c_str());
else {
cerr << "WARNING: Sequencer: Unkown configuration parameter: " << argv[i] << endl;
assert(false);
}
}
assert(m_max_outstanding_requests > 0);
assert(m_deadlock_threshold > 0);
assert(m_version > -1);
assert(m_instCache_ptr != NULL);
assert(m_dataCache_ptr != NULL);
assert(m_controller != NULL);
}
Sequencer::~Sequencer() {
}
void Sequencer::wakeup() {
// Check for deadlock of any of the requests
Time current_time = g_eventQueue_ptr->getTime();
// Check across all outstanding requests
int total_outstanding = 0;
Vector<Address> keys = m_readRequestTable.keys();
for (int i=0; i<keys.size(); i++) {
SequencerRequest* request = m_readRequestTable.lookup(keys[i]);
if (current_time - request->issue_time >= m_deadlock_threshold) {
WARN_MSG("Possible Deadlock detected");
WARN_EXPR(request);
WARN_EXPR(m_version);
WARN_EXPR(keys.size());
WARN_EXPR(current_time);
WARN_EXPR(request->issue_time);
WARN_EXPR(current_time - request->issue_time);
ERROR_MSG("Aborting");
}
}
keys = m_writeRequestTable.keys();
for (int i=0; i<keys.size(); i++) {
SequencerRequest* request = m_writeRequestTable.lookup(keys[i]);
if (current_time - request->issue_time >= m_deadlock_threshold) {
WARN_MSG("Possible Deadlock detected");
WARN_EXPR(request);
WARN_EXPR(m_version);
WARN_EXPR(current_time);
WARN_EXPR(request->issue_time);
WARN_EXPR(current_time - request->issue_time);
WARN_EXPR(keys.size());
ERROR_MSG("Aborting");
}
}
total_outstanding += m_writeRequestTable.size() + m_readRequestTable.size();
assert(m_outstanding_count == total_outstanding);
if (m_outstanding_count > 0) { // If there are still outstanding requests, keep checking
g_eventQueue_ptr->scheduleEvent(this, m_deadlock_threshold);
} else {
m_deadlock_check_scheduled = false;
}
}
void Sequencer::printProgress(ostream& out) const{
/*
int total_demand = 0;
out << "Sequencer Stats Version " << m_version << endl;
out << "Current time = " << g_eventQueue_ptr->getTime() << endl;
out << "---------------" << endl;
out << "outstanding requests" << endl;
Vector<Address> rkeys = m_readRequestTable.keys();
int read_size = rkeys.size();
out << "proc " << m_version << " Read Requests = " << read_size << endl;
// print the request table
for(int i=0; i < read_size; ++i){
SequencerRequest * request = m_readRequestTable.lookup(rkeys[i]);
out << "\tRequest[ " << i << " ] = " << request->type << " Address " << rkeys[i] << " Posted " << request->issue_time << " PF " << PrefetchBit_No << endl;
total_demand++;
}
Vector<Address> wkeys = m_writeRequestTable.keys();
int write_size = wkeys.size();
out << "proc " << m_version << " Write Requests = " << write_size << endl;
// print the request table
for(int i=0; i < write_size; ++i){
CacheMsg & request = m_writeRequestTable.lookup(wkeys[i]);
out << "\tRequest[ " << i << " ] = " << request.getType() << " Address " << wkeys[i] << " Posted " << request.getTime() << " PF " << request.getPrefetch() << endl;
if( request.getPrefetch() == PrefetchBit_No ){
total_demand++;
}
}
out << endl;
out << "Total Number Outstanding: " << m_outstanding_count << endl;
out << "Total Number Demand : " << total_demand << endl;
out << "Total Number Prefetches : " << m_outstanding_count - total_demand << endl;
out << endl;
out << endl;
*/
}
void Sequencer::printConfig(ostream& out) const {
out << "Seqeuncer config: " << m_name << endl;
out << " controller: " << m_controller->getName() << endl;
out << " version: " << m_version << endl;
out << " max_outstanding_requests: " << m_max_outstanding_requests << endl;
out << " deadlock_threshold: " << m_deadlock_threshold << endl;
}
// Insert the request on the correct request table. Return true if
// the entry was already present.
bool Sequencer::insertRequest(SequencerRequest* request) {
int total_outstanding = m_writeRequestTable.size() + m_readRequestTable.size();
assert(m_outstanding_count == total_outstanding);
// See if we should schedule a deadlock check
if (m_deadlock_check_scheduled == false) {
g_eventQueue_ptr->scheduleEvent(this, m_deadlock_threshold);
m_deadlock_check_scheduled = true;
}
Address line_addr(request->ruby_request.paddr);
line_addr.makeLineAddress();
if ((request->ruby_request.type == RubyRequestType_ST) ||
(request->ruby_request.type == RubyRequestType_RMW_Read) ||
(request->ruby_request.type == RubyRequestType_RMW_Write) ||
(request->ruby_request.type == RubyRequestType_Locked_Read) ||
(request->ruby_request.type == RubyRequestType_Locked_Write)) {
if (m_writeRequestTable.exist(line_addr)) {
m_writeRequestTable.lookup(line_addr) = request;
// return true;
assert(0); // drh5: isn't this an error? do you lose the initial request?
}
m_writeRequestTable.allocate(line_addr);
m_writeRequestTable.lookup(line_addr) = request;
m_outstanding_count++;
} else {
if (m_readRequestTable.exist(line_addr)) {
m_readRequestTable.lookup(line_addr) = request;
// return true;
assert(0); // drh5: isn't this an error? do you lose the initial request?
}
m_readRequestTable.allocate(line_addr);
m_readRequestTable.lookup(line_addr) = request;
m_outstanding_count++;
}
g_system_ptr->getProfiler()->sequencerRequests(m_outstanding_count);
total_outstanding = m_writeRequestTable.size() + m_readRequestTable.size();
assert(m_outstanding_count == total_outstanding);
return false;
}
void Sequencer::removeRequest(SequencerRequest* srequest) {
assert(m_outstanding_count == m_writeRequestTable.size() + m_readRequestTable.size());
const RubyRequest & ruby_request = srequest->ruby_request;
Address line_addr(ruby_request.paddr);
line_addr.makeLineAddress();
if ((ruby_request.type == RubyRequestType_ST) ||
(ruby_request.type == RubyRequestType_RMW_Read) ||
(ruby_request.type == RubyRequestType_RMW_Write) ||
(ruby_request.type == RubyRequestType_Locked_Read) ||
(ruby_request.type == RubyRequestType_Locked_Write)) {
m_writeRequestTable.deallocate(line_addr);
} else {
m_readRequestTable.deallocate(line_addr);
}
m_outstanding_count--;
assert(m_outstanding_count == m_writeRequestTable.size() + m_readRequestTable.size());
}
void Sequencer::writeCallback(const Address& address, DataBlock& data) {
assert(address == line_address(address));
assert(m_writeRequestTable.exist(line_address(address)));
SequencerRequest* request = m_writeRequestTable.lookup(address);
removeRequest(request);
assert((request->ruby_request.type == RubyRequestType_ST) ||
(request->ruby_request.type == RubyRequestType_RMW_Read) ||
(request->ruby_request.type == RubyRequestType_RMW_Write) ||
(request->ruby_request.type == RubyRequestType_Locked_Read) ||
(request->ruby_request.type == RubyRequestType_Locked_Write));
// POLINA: the assumption is that atomics are only on data cache and not instruction cache
if (request->ruby_request.type == RubyRequestType_Locked_Read) {
m_dataCache_ptr->setLocked(address, m_version);
}
hitCallback(request, data);
}
void Sequencer::readCallback(const Address& address, DataBlock& data) {
assert(address == line_address(address));
assert(m_readRequestTable.exist(line_address(address)));
SequencerRequest* request = m_readRequestTable.lookup(address);
removeRequest(request);
assert((request->ruby_request.type == RubyRequestType_LD) ||
(request->ruby_request.type == RubyRequestType_RMW_Read) ||
(request->ruby_request.type == RubyRequestType_IFETCH));
hitCallback(request, data);
}
void Sequencer::hitCallback(SequencerRequest* srequest, DataBlock& data) {
const RubyRequest & ruby_request = srequest->ruby_request;
Address request_address(ruby_request.paddr);
Address request_line_address(ruby_request.paddr);
request_line_address.makeLineAddress();
RubyRequestType type = ruby_request.type;
Time issued_time = srequest->issue_time;
// Set this cache entry to the most recently used
if (type == RubyRequestType_IFETCH) {
if (m_instCache_ptr->isTagPresent(request_line_address) )
m_instCache_ptr->setMRU(request_line_address);
} else {
if (m_dataCache_ptr->isTagPresent(request_line_address) )
m_dataCache_ptr->setMRU(request_line_address);
}
assert(g_eventQueue_ptr->getTime() >= issued_time);
Time miss_latency = g_eventQueue_ptr->getTime() - issued_time;
// Profile the miss latency for all non-zero demand misses
if (miss_latency != 0) {
g_system_ptr->getProfiler()->missLatency(miss_latency, type);
if (Debug::getProtocolTrace()) {
g_system_ptr->getProfiler()->profileTransition("Seq", m_version, Address(ruby_request.paddr),
"", "Done", "", int_to_string(miss_latency)+" cycles");
}
}
/*
if (request.getPrefetch() == PrefetchBit_Yes) {
return; // Ignore the prefetch
}
*/
// update the data
if (ruby_request.data != NULL) {
if ((type == RubyRequestType_LD) ||
(type == RubyRequestType_IFETCH) ||
(type == RubyRequestType_RMW_Read)) {
memcpy(ruby_request.data, data.getData(request_address.getOffset(), ruby_request.len), ruby_request.len);
} else {
data.setData(ruby_request.data, request_address.getOffset(), ruby_request.len);
}
}
m_hit_callback(srequest->id);
delete srequest;
}
// Returns true if the sequencer already has a load or store outstanding
bool Sequencer::isReady(const RubyRequest& request) const {
// POLINA: check if we are currently flushing the write buffer, if so Ruby is returned as not ready
// to simulate stalling of the front-end
// Do we stall all the sequencers? If it is atomic instruction - yes!
if (m_outstanding_count >= m_max_outstanding_requests) {
return false;
}
if( m_writeRequestTable.exist(line_address(Address(request.paddr))) ||
m_readRequestTable.exist(line_address(Address(request.paddr))) ){
//cout << "OUTSTANDING REQUEST EXISTS " << p << " VER " << m_version << endl;
//printProgress(cout);
return false;
}
return true;
}
bool Sequencer::empty() const {
return (m_writeRequestTable.size() == 0) && (m_readRequestTable.size() == 0);
}
int64_t Sequencer::makeRequest(const RubyRequest & request)
{
assert(Address(request.paddr).getOffset() + request.len <= RubySystem::getBlockSizeBytes());
if (isReady(request)) {
int64_t id = makeUniqueRequestID();
SequencerRequest *srequest = new SequencerRequest(request, id, g_eventQueue_ptr->getTime());
bool found = insertRequest(srequest);
if (!found)
if (request.type == RubyRequestType_Locked_Write) {
// NOTE: it is OK to check the locked flag here as the mandatory queue will be checked first
// ensuring that nothing comes between checking the flag and servicing the store
if (!m_dataCache_ptr->isLocked(line_address(Address(request.paddr)), m_version)) {
return LLSC_FAIL;
}
else {
m_dataCache_ptr->clearLocked(line_address(Address(request.paddr)));
}
}
issueRequest(request);
// TODO: issue hardware prefetches here
return id;
}
else {
return -1;
}
}
void Sequencer::issueRequest(const RubyRequest& request) {
// TODO: get rid of CacheMsg, CacheRequestType, and AccessModeTYpe, & have SLICC use RubyRequest and subtypes natively
CacheRequestType ctype;
switch(request.type) {
case RubyRequestType_IFETCH:
ctype = CacheRequestType_IFETCH;
break;
case RubyRequestType_LD:
ctype = CacheRequestType_LD;
break;
case RubyRequestType_ST:
ctype = CacheRequestType_ST;
break;
case RubyRequestType_Locked_Read:
ctype = CacheRequestType_ST;
break;
case RubyRequestType_Locked_Write:
ctype = CacheRequestType_ST;
break;
case RubyRequestType_RMW_Read:
ctype = CacheRequestType_ATOMIC;
break;
case RubyRequestType_RMW_Write:
ctype = CacheRequestType_ATOMIC;
break;
default:
assert(0);
}
AccessModeType amtype;
switch(request.access_mode){
case RubyAccessMode_User:
amtype = AccessModeType_UserMode;
break;
case RubyAccessMode_Supervisor:
amtype = AccessModeType_SupervisorMode;
break;
case RubyAccessMode_Device:
amtype = AccessModeType_UserMode;
break;
default:
assert(0);
}
Address line_addr(request.paddr);
line_addr.makeLineAddress();
CacheMsg msg(line_addr, Address(request.paddr), ctype, Address(request.pc), amtype, request.len, PrefetchBit_No);
if (Debug::getProtocolTrace()) {
g_system_ptr->getProfiler()->profileTransition("Seq", m_version, Address(request.paddr),
"", "Begin", "", RubyRequestType_to_string(request.type));
}
if (g_system_ptr->getTracer()->traceEnabled()) {
g_system_ptr->getTracer()->traceRequest(m_name, line_addr, Address(request.pc),
request.type, g_eventQueue_ptr->getTime());
}
Time latency = 0; // initialzed to an null value
if (request.type == RubyRequestType_IFETCH)
latency = m_instCache_ptr->getLatency();
else
latency = m_dataCache_ptr->getLatency();
// Send the message to the cache controller
assert(latency > 0);
m_mandatory_q_ptr->enqueue(msg, latency);
}
/*
bool Sequencer::tryCacheAccess(const Address& addr, CacheRequestType type,
AccessModeType access_mode,
int size, DataBlock*& data_ptr) {
if (type == CacheRequestType_IFETCH) {
return m_instCache_ptr->tryCacheAccess(line_address(addr), type, data_ptr);
} else {
return m_dataCache_ptr->tryCacheAccess(line_address(addr), type, data_ptr);
}
}
*/
void Sequencer::print(ostream& out) const {
out << "[Sequencer: " << m_version
<< ", outstanding requests: " << m_outstanding_count;
out << ", read request table: " << m_readRequestTable
<< ", write request table: " << m_writeRequestTable;
out << "]";
}
// this can be called from setState whenever coherence permissions are upgraded
// when invoked, coherence violations will be checked for the given block
void Sequencer::checkCoherence(const Address& addr) {
#ifdef CHECK_COHERENCE
g_system_ptr->checkGlobalCoherenceInvariant(addr);
#endif
}