Files
gem5/src/mem/ruby/slicc_interface/AbstractController.hh
Marleson Graf 63d110fb7a mem-ruby,sim-se: Support Maybe_Stale in functional reads
Functional reads can be satisfied by one of the following, in order:
1. Main memory (when the data is not present in the cache hierarchy);
2. Valid data block in cache;
3. Valid data block in coherence message;
4. Valid data block marked as Maybe_Stale;

Number 4 is not handled by the current implementation. A Maybe_Stale
block can be either truly stale or actually valid. When it is stale,
the memory read will be satisfied by either number 2 or number 3. When
it is valid, there will be no coherence message with valid data inside,
and the Maybe_Stale block will transition to a valid state after
receiving some kind of acknowledgement.

The main challenge to handle number 4 is how to know from which
Maybe_Stale block the data should be read from. For instance, in a two
level cache hierarchy, we might have a block marked as Maybe_Stale in
both L1 and L2. In this case, we should prioritize the cache controller
that is closest to the CPU. To define this priority, a new virtual
function 'functionalReadPriority' was added to the AbstractController
class.

Change-Id: I4774cd01aab7bb9ca53694cd9dc4f9416a8e4025
2024-11-18 00:22:36 -08:00

530 lines
20 KiB
C++

/*
* Copyright (c) 2017,2019-2023 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.
*
* Copyright (c) 2009-2014 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.
*/
#ifndef __MEM_RUBY_SLICC_INTERFACE_ABSTRACTCONTROLLER_HH__
#define __MEM_RUBY_SLICC_INTERFACE_ABSTRACTCONTROLLER_HH__
#include <exception>
#include <iostream>
#include <string>
#include <unordered_map>
#include "base/addr_range.hh"
#include "base/addr_range_map.hh"
#include "base/callback.hh"
#include "mem/packet.hh"
#include "mem/qport.hh"
#include "mem/ruby/common/Address.hh"
#include "mem/ruby/common/Consumer.hh"
#include "mem/ruby/common/DataBlock.hh"
#include "mem/ruby/common/Histogram.hh"
#include "mem/ruby/common/MachineID.hh"
#include "mem/ruby/network/MessageBuffer.hh"
#include "mem/ruby/protocol/AccessPermission.hh"
#include "mem/ruby/system/CacheRecorder.hh"
#include "params/RubyController.hh"
#include "sim/clocked_object.hh"
#include "sim/eventq.hh"
namespace gem5
{
namespace ruby
{
class Network;
class GPUCoalescer;
class DMASequencer;
class RubySystem;
// used to communicate that an in_port peeked the wrong message type
class RejectException: public std::exception
{
virtual const char* what() const throw()
{ return "Port rejected message based on type"; }
};
class AbstractController : public ClockedObject, public Consumer
{
public:
PARAMS(RubyController);
AbstractController(const Params &p);
void init();
NodeID getVersion() const { return m_machineID.getNum(); }
MachineType getType() const { return m_machineID.getType(); }
void initNetworkPtr(Network* net_ptr) { m_net_ptr = net_ptr; }
// return instance name
void blockOnQueue(Addr, MessageBuffer*);
bool isBlocked(Addr) const;
void unblock(Addr);
bool isBlocked(Addr);
virtual MessageBuffer* getMandatoryQueue() const = 0;
virtual MessageBuffer* getMemReqQueue() const = 0;
virtual MessageBuffer* getMemRespQueue() const = 0;
// That function must be called by controller when dequeuing mem resp queue
// for memory controller to receive the retry request in time
void memRespQueueDequeued();
// Or that function can be called to perform both dequeue and notification
// at once.
void dequeueMemRespQueue();
virtual AccessPermission getAccessPermission(const Addr &addr) = 0;
virtual void print(std::ostream & out) const = 0;
virtual void wakeup() = 0;
virtual void resetStats() = 0;
virtual void regStats();
virtual void recordCacheTrace(int cntrl, CacheRecorder* tr) = 0;
virtual Sequencer* getCPUSequencer() const = 0;
virtual DMASequencer* getDMASequencer() const = 0;
virtual GPUCoalescer* getGPUCoalescer() const = 0;
// This latency is used by the sequencer when enqueueing requests.
// Different latencies may be used depending on the request type.
// This is the hit latency unless the top-level cache controller
// introduces additional cycles in the response path.
virtual Cycles mandatoryQueueLatency(const RubyRequestType& param_type)
{ return m_mandatory_queue_latency; }
//! These functions are used by ruby system to read/write the data blocks
//! that exist with in the controller.
virtual bool functionalReadBuffers(PacketPtr&) = 0;
virtual void functionalRead(const Addr &addr, PacketPtr)
{ panic("functionalRead(Addr,PacketPtr) not implemented"); }
/**
* Returns the priority used by functional reads when deciding from which
* controller to read a Maybe_Stale data block.
* Lower positive values have higher priority, negative values are ignored.
*
* @return the controller's priority
*/
virtual int functionalReadPriority() { return -1; }
//! Functional read that reads only blocks not present in the mask.
//! Return number of bytes read.
virtual bool functionalReadBuffers(PacketPtr&, WriteMask &mask) = 0;
virtual void functionalRead(const Addr &addr, PacketPtr pkt,
WriteMask &mask)
{ panic("functionalRead(Addr,PacketPtr,WriteMask) not implemented"); }
void functionalMemoryRead(PacketPtr);
//! The return value indicates the number of messages written with the
//! data from the packet.
virtual int functionalWriteBuffers(PacketPtr&) = 0;
virtual int functionalWrite(const Addr &addr, PacketPtr) = 0;
int functionalMemoryWrite(PacketPtr);
//! Function for enqueuing a prefetch request
virtual void enqueuePrefetch(const Addr &, const RubyRequestType&)
{ fatal("Prefetches not implemented!");}
//! Notifies controller of a request coalesced at the sequencer.
//! By default, it does nothing. Behavior is protocol-specific
virtual void notifyCoalesced(const Addr& addr,
const RubyRequestType& type,
const RequestPtr& req,
const DataBlock& data_blk,
const bool& was_miss)
{ }
//! Function for collating statistics from all the controllers of this
//! particular type. This function should only be called from the
//! version 0 of this controller type.
virtual void collateStats()
{fatal("collateStats() should be overridden!");}
//! Initialize the message buffers.
virtual void initNetQueues() = 0;
/** A function used to return the port associated with this bus object. */
Port &getPort(const std::string &if_name,
PortID idx=InvalidPortID);
bool recvTimingResp(PacketPtr pkt);
Tick recvAtomic(PacketPtr pkt);
const AddrRangeList &getAddrRanges() const { return addrRanges; }
public:
MachineID getMachineID() const { return m_machineID; }
RequestorID getRequestorId() const { return m_id; }
statistics::Histogram& getDelayHist() { return stats.delayHistogram; }
statistics::Histogram& getDelayVCHist(uint32_t index)
{ return *(stats.delayVCHistogram[index]); }
bool respondsTo(Addr addr)
{
for (auto &range: addrRanges)
if (range.contains(addr)) return true;
return false;
}
/**
* Map an address to the correct MachineID
*
* This function querries the network for the NodeID of the
* destination for a given request using its address and the type
* of the destination. For example for a request with a given
* address to a directory it will return the MachineID of the
* authorative directory.
*
* @param the destination address
* @param the type of the destination
* @return the MachineID of the destination
*/
MachineID mapAddressToMachine(Addr addr, MachineType mtype) const;
/**
* Maps an address to the correct dowstream MachineID (i.e. the component
* in the next level of the cache hierarchy towards memory)
*
* This function uses the local list of possible destinations instead of
* querying the network.
*
* @param the destination address
* @param the type of the destination (optional)
* @return the MachineID of the destination
*/
MachineID mapAddressToDownstreamMachine(Addr addr,
MachineType mtype = MachineType_NUM) const;
/** List of downstream destinations (towards memory) */
const NetDest& allDownstreamDest() const { return downstreamDestinations; }
/** List of upstream destinations (towards the CPU) */
const NetDest& allUpstreamDest() const { return upstreamDestinations; }
// Helper methods for commonly used functions called in common/address.hh
Addr getOffset(Addr addr) const;
Addr makeLineAddress(Addr addr) const;
std::string printAddress(Addr addr) const;
protected:
//! Profiles original cache requests including PUTs
void profileRequest(const std::string &request);
//! Profiles the delay associated with messages.
void profileMsgDelay(uint32_t virtualNetwork, Cycles delay);
// Tracks outstanding transactions for latency profiling
struct TransMapPair { unsigned transaction; unsigned state; Tick time; };
std::unordered_map<Addr, TransMapPair> m_inTransAddressed;
std::unordered_map<Addr, TransMapPair> m_outTransAddressed;
std::unordered_map<Addr, TransMapPair> m_inTransUnaddressed;
std::unordered_map<Addr, TransMapPair> m_outTransUnaddressed;
/**
* Profiles an event that initiates a protocol transactions for a specific
* line (e.g. events triggered by incoming request messages).
* A histogram with the latency of the transactions is generated for
* all combinations of trigger event, initial state, and final state.
* This function also supports "unaddressed" transactions,
* those not associated with an address in memory but
* instead associated with a unique ID.
*
* @param addr address of the line, or unique transaction ID
* @param type event that started the transaction
* @param initialState state of the line before the transaction
* @param isAddressed is addr a line address or a unique ID
*/
template<typename EventType, typename StateType>
void incomingTransactionStart(Addr addr,
EventType type, StateType initialState, bool retried,
bool isAddressed=true)
{
auto& m_inTrans =
isAddressed ? m_inTransAddressed : m_inTransUnaddressed;
assert(m_inTrans.find(addr) == m_inTrans.end());
m_inTrans[addr] = {type, initialState, curTick()};
if (retried)
++(*stats.inTransRetryCnt[type]);
}
/**
* Profiles an event that ends a transaction.
* This function also supports "unaddressed" transactions,
* those not associated with an address in memory but
* instead associated with a unique ID.
*
* @param addr address or unique ID with an outstanding transaction
* @param finalState state of the line after the transaction
* @param isAddressed is addr a line address or a unique ID
*/
template<typename StateType>
void incomingTransactionEnd(Addr addr, StateType finalState,
bool isAddressed=true)
{
auto& m_inTrans =
isAddressed ? m_inTransAddressed : m_inTransUnaddressed;
auto iter = m_inTrans.find(addr);
assert(iter != m_inTrans.end());
auto &trans = iter->second;
auto stat_iter_ev = stats.inTransStateChanges.find(trans.transaction);
gem5_assert(stat_iter_ev != stats.inTransStateChanges.end(),
"%s: event type=%d not marked as in_trans in SLICC",
name(), trans.transaction);
auto stat_iter_state = stat_iter_ev->second.find(trans.state);
gem5_assert(stat_iter_state != stat_iter_ev->second.end(),
"%s: event type=%d has no transition from state=%d",
name(), trans.transaction, trans.state);
++(*stat_iter_state->second[(unsigned)finalState]);
stats.inTransLatHist[iter->second.transaction]->sample(
ticksToCycles(curTick() - trans.time));
m_inTrans.erase(iter);
}
/**
* Profiles an event that initiates a transaction in a peer controller
* (e.g. an event that sends a request message)
* This function also supports "unaddressed" transactions,
* those not associated with an address in memory but
* instead associated with a unique ID.
*
* @param addr address of the line or a unique transaction ID
* @param type event that started the transaction
* @param isAddressed is addr a line address or a unique ID
*/
template<typename EventType>
void outgoingTransactionStart(Addr addr, EventType type,
bool isAddressed=true)
{
auto& m_outTrans =
isAddressed ? m_outTransAddressed : m_outTransUnaddressed;
assert(m_outTrans.find(addr) == m_outTrans.end());
m_outTrans[addr] = {type, 0, curTick()};
}
/**
* Profiles the end of an outgoing transaction.
* (e.g. receiving the response for a requests)
* This function also supports "unaddressed" transactions,
* those not associated with an address in memory but
* instead associated with a unique ID.
*
* @param addr address of the line with an outstanding transaction
* @param isAddressed is addr a line address or a unique ID
*/
void outgoingTransactionEnd(Addr addr, bool retried,
bool isAddressed=true)
{
auto& m_outTrans =
isAddressed ? m_outTransAddressed : m_outTransUnaddressed;
auto iter = m_outTrans.find(addr);
assert(iter != m_outTrans.end());
auto &trans = iter->second;
auto stat_iter = stats.outTransLatHist.find(trans.transaction);
gem5_assert(stat_iter != stats.outTransLatHist.end(),
"%s: event type=%d not marked as out_trans in SLICC",
name(), trans.transaction);
stat_iter->second->sample(
ticksToCycles(curTick() - trans.time));
if (retried)
++(*stats.outTransRetryCnt[trans.transaction]);
m_outTrans.erase(iter);
}
void stallBuffer(MessageBuffer* buf, Addr addr);
void wakeUpBuffer(MessageBuffer* buf, Addr addr);
void wakeUpBuffers(Addr addr);
void wakeUpAllBuffers(Addr addr);
void wakeUpAllBuffers();
bool serviceMemoryQueue();
/**
* Functions needed by CacheAccessor. These are implemented in SLICC,
* thus the const& for all args to match the generated code.
*/
virtual bool inCache(const Addr &addr, const bool &is_secure)
{ fatal("inCache: prefetching not supported"); return false; }
virtual bool hasBeenPrefetched(const Addr &addr, const bool &is_secure)
{ fatal("hasBeenPrefetched: prefetching not supported"); return false; }
virtual bool hasBeenPrefetched(const Addr &addr, const bool &is_secure,
const RequestorID &requestor)
{ fatal("hasBeenPrefetched: prefetching not supported"); return false; }
virtual bool inMissQueue(const Addr &addr, const bool &is_secure)
{ fatal("inMissQueue: prefetching not supported"); return false; }
virtual bool coalesce()
{ fatal("coalesce: prefetching not supported"); return false; }
friend class RubyPrefetcherProxy;
protected:
const NodeID m_version;
MachineID m_machineID;
const NodeID m_clusterID;
// RequestorID used by some components of gem5.
const RequestorID m_id;
Network *m_net_ptr;
bool m_is_blocking;
std::map<Addr, MessageBuffer*> m_block_map;
typedef std::vector<MessageBuffer*> MsgVecType;
typedef std::set<MessageBuffer*> MsgBufType;
typedef std::map<Addr, MsgVecType* > WaitingBufType;
WaitingBufType m_waiting_buffers;
unsigned int m_in_ports;
unsigned int m_cur_in_port;
const int m_number_of_TBEs;
const int m_transitions_per_cycle;
const unsigned int m_buffer_size;
Cycles m_recycle_latency;
const Cycles m_mandatory_queue_latency;
bool m_waiting_mem_retry;
bool m_mem_ctrl_waiting_retry;
/**
* Port that forwards requests and receives responses from the
* memory controller.
*/
class MemoryPort : public RequestPort
{
private:
// Controller that operates this port.
AbstractController *controller;
public:
MemoryPort(const std::string &_name, AbstractController *_controller,
PortID id = InvalidPortID);
protected:
// Function for receiving a timing response from the peer port.
// Currently the pkt is handed to the coherence controller
// associated with this port.
bool recvTimingResp(PacketPtr pkt);
void recvReqRetry();
};
/* Request port to the memory controller. */
MemoryPort memoryPort;
// State that is stored in packets sent to the memory controller.
struct SenderState : public Packet::SenderState
{
// Id of the machine from which the request originated.
MachineID id;
SenderState(MachineID _id) : id(_id)
{}
};
RubySystem *m_ruby_system = nullptr;
// Formerly in RubySlicc_ComponentMapping.hh. Moved here to access
// RubySystem pointer.
NetDest broadcast(MachineType type);
int machineCount(MachineType machType);
private:
/** The address range to which the controller responds on the CPU side. */
const AddrRangeList addrRanges;
std::unordered_map<MachineType, AddrRangeMap<MachineID, 3>>
downstreamAddrMap;
NetDest downstreamDestinations;
NetDest upstreamDestinations;
void sendRetryRespToMem();
MemberEventWrapper<&AbstractController::sendRetryRespToMem> mRetryRespEvent;
public:
struct ControllerStats : public statistics::Group
{
ControllerStats(statistics::Group *parent);
// Initialized by the SLICC compiler for all events with the
// "in_trans" property.
// Only histograms with samples will appear in the stats
std::unordered_map<unsigned, statistics::Histogram*> inTransLatHist;
std::unordered_map<unsigned, statistics::Scalar*> inTransRetryCnt;
// Initialized by the SLICC compiler for all combinations of events
// with the "in_trans" property, potential initial states, and
// potential final states. Potential initial states are states that
// appear in transitions triggered by that event. Currently all states
// are considered as potential final states.
std::unordered_map<unsigned, std::unordered_map<unsigned,
std::vector<statistics::Scalar*>>> inTransStateChanges;
// Initialized by the SLICC compiler for all events with the
// "out_trans" property.
// Only histograms with samples will appear in the stats.
std::unordered_map<unsigned, statistics::Histogram*> outTransLatHist;
std::unordered_map<unsigned, statistics::Scalar*>
outTransRetryCnt;
//! Counter for the number of cycles when the transitions carried out
//! were equal to the maximum allowed
statistics::Scalar fullyBusyCycles;
//! Histogram for profiling delay for the messages this controller
//! cares for
statistics::Histogram delayHistogram;
std::vector<statistics::Histogram *> delayVCHistogram;
} stats;
};
} // namespace ruby
} // namespace gem5
#endif // __MEM_RUBY_SLICC_INTERFACE_ABSTRACTCONTROLLER_HH__