gem5  v22.1.0.0
Classes | Public Types | Public Member Functions | Protected Attributes | Private Attributes | List of all members
gem5::GPUDispatcher Class Reference

#include <dispatcher.hh>

Inheritance diagram for gem5::GPUDispatcher:
gem5::SimObject gem5::EventManager gem5::Serializable gem5::Drainable gem5::statistics::Group gem5::Named

Classes

struct  GPUDispatcherStats
 

Public Types

typedef GPUDispatcherParams Params
 
- Public Types inherited from gem5::SimObject
typedef SimObjectParams Params
 

Public Member Functions

 GPUDispatcher (const Params &p)
 
 ~GPUDispatcher ()
 
void serialize (CheckpointOut &cp) const override
 Serialize an object. More...
 
void unserialize (CheckpointIn &cp) override
 Unserialize an object. More...
 
void setCommandProcessor (GPUCommandProcessor *gpu_cmd_proc)
 
void setShader (Shader *new_shader)
 
void exec ()
 
bool isReachingKernelEnd (Wavefront *wf)
 
void updateInvCounter (int kern_id, int val=-1)
 update the counter of oustanding inv requests for the kernel kern_id: kernel id val: +1/-1, increment or decrement the counter (default: -1) More...
 
bool updateWbCounter (int kern_id, int val=-1)
 update the counter of oustanding wb requests for the kernel kern_id: kernel id val: +1/-1, increment or decrement the counter (default: -1) More...
 
int getOutstandingWbs (int kern_id)
 get kernel's outstanding cache writeback requests More...
 
void notifyWgCompl (Wavefront *wf)
 When an end program instruction detects that the last WF in a WG has completed it will call this method on the dispatcher. More...
 
void scheduleDispatch ()
 
void dispatch (HSAQueueEntry *task)
 After all relevant HSA data structures have been traversed/extracted from memory by the CP, dispatch() is called on the dispatcher. More...
 
HSAQueueEntryhsaTask (int disp_id)
 
- Public Member Functions inherited from gem5::SimObject
const Paramsparams () const
 
 SimObject (const Params &p)
 
virtual ~SimObject ()
 
virtual void init ()
 init() is called after all C++ SimObjects have been created and all ports are connected. More...
 
virtual void loadState (CheckpointIn &cp)
 loadState() is called on each SimObject when restoring from a checkpoint. More...
 
virtual void initState ()
 initState() is called on each SimObject when not restoring from a checkpoint. More...
 
virtual void regProbePoints ()
 Register probe points for this object. More...
 
virtual void regProbeListeners ()
 Register probe listeners for this object. More...
 
ProbeManagergetProbeManager ()
 Get the probe manager for this object. More...
 
virtual PortgetPort (const std::string &if_name, PortID idx=InvalidPortID)
 Get a port with a given name and index. More...
 
virtual void startup ()
 startup() is the final initialization call before simulation. More...
 
DrainState drain () override
 Provide a default implementation of the drain interface for objects that don't need draining. More...
 
virtual void memWriteback ()
 Write back dirty buffers to memory using functional writes. More...
 
virtual void memInvalidate ()
 Invalidate the contents of memory buffers. More...
 
void serialize (CheckpointOut &cp) const override
 Serialize an object. More...
 
void unserialize (CheckpointIn &cp) override
 Unserialize an object. More...
 
- Public Member Functions inherited from gem5::EventManager
EventQueueeventQueue () const
 
void schedule (Event &event, Tick when)
 
void deschedule (Event &event)
 
void reschedule (Event &event, Tick when, bool always=false)
 
void schedule (Event *event, Tick when)
 
void deschedule (Event *event)
 
void reschedule (Event *event, Tick when, bool always=false)
 
void wakeupEventQueue (Tick when=(Tick) -1)
 This function is not needed by the usual gem5 event loop but may be necessary in derived EventQueues which host gem5 on other schedulers. More...
 
void setCurTick (Tick newVal)
 
 EventManager (EventManager &em)
 Event manger manages events in the event queue. More...
 
 EventManager (EventManager *em)
 
 EventManager (EventQueue *eq)
 
- Public Member Functions inherited from gem5::Serializable
 Serializable ()
 
virtual ~Serializable ()
 
void serializeSection (CheckpointOut &cp, const char *name) const
 Serialize an object into a new section. More...
 
void serializeSection (CheckpointOut &cp, const std::string &name) const
 
void unserializeSection (CheckpointIn &cp, const char *name)
 Unserialize an a child object. More...
 
void unserializeSection (CheckpointIn &cp, const std::string &name)
 
- Public Member Functions inherited from gem5::Drainable
DrainState drainState () const
 Return the current drain state of an object. More...
 
virtual void notifyFork ()
 Notify a child process of a fork. More...
 
- Public Member Functions inherited from gem5::statistics::Group
 Group (Group *parent, const char *name=nullptr)
 Construct a new statistics group. More...
 
virtual ~Group ()
 
virtual void regStats ()
 Callback to set stat parameters. More...
 
virtual void resetStats ()
 Callback to reset stats. More...
 
virtual void preDumpStats ()
 Callback before stats are dumped. More...
 
void addStat (statistics::Info *info)
 Register a stat with this group. More...
 
const std::map< std::string, Group * > & getStatGroups () const
 Get all child groups associated with this object. More...
 
const std::vector< Info * > & getStats () const
 Get all stats associated with this object. More...
 
void addStatGroup (const char *name, Group *block)
 Add a stat block as a child of this block. More...
 
const InforesolveStat (std::string name) const
 Resolve a stat by its name within this group. More...
 
void mergeStatGroup (Group *block)
 Merge the contents (stats & children) of a block to this block. More...
 
 Group ()=delete
 
 Group (const Group &)=delete
 
Groupoperator= (const Group &)=delete
 
- Public Member Functions inherited from gem5::Named
 Named (const std::string &name_)
 
virtual ~Named ()=default
 
virtual std::string name () const
 

Protected Attributes

gem5::GPUDispatcher::GPUDispatcherStats stats
 
- Protected Attributes inherited from gem5::SimObject
const SimObjectParams & _params
 Cached copy of the object parameters. More...
 
- Protected Attributes inherited from gem5::EventManager
EventQueueeventq
 A pointer to this object's event queue. More...
 

Private Attributes

Shadershader
 
GPUCommandProcessorgpuCmdProc
 
EventFunctionWrapper tickEvent
 
std::unordered_map< int, HSAQueueEntry * > hsaQueueEntries
 
std::queue< int > execIds
 
std::queue< int > doneIds
 
bool dispatchActive
 

Additional Inherited Members

- Static Public Member Functions inherited from gem5::SimObject
static void serializeAll (const std::string &cpt_dir)
 Create a checkpoint by serializing all SimObjects in the system. More...
 
static SimObjectfind (const char *name)
 Find the SimObject with the given name and return a pointer to it. More...
 
static void setSimObjectResolver (SimObjectResolver *resolver)
 There is a single object name resolver, and it is only set when simulation is restoring from checkpoints. More...
 
static SimObjectResolvergetSimObjectResolver ()
 There is a single object name resolver, and it is only set when simulation is restoring from checkpoints. More...
 
- Static Public Member Functions inherited from gem5::Serializable
static const std::string & currentSection ()
 Gets the fully-qualified name of the active section. More...
 
static void generateCheckpointOut (const std::string &cpt_dir, std::ofstream &outstream)
 Generate a checkpoint file so that the serialization can be routed to it. More...
 
- Protected Member Functions inherited from gem5::Drainable
 Drainable ()
 
virtual ~Drainable ()
 
virtual void drainResume ()
 Resume execution after a successful drain. More...
 
void signalDrainDone () const
 Signal that an object is drained. More...
 

Detailed Description

Definition at line 62 of file dispatcher.hh.

Member Typedef Documentation

◆ Params

typedef GPUDispatcherParams gem5::GPUDispatcher::Params

Definition at line 65 of file dispatcher.hh.

Constructor & Destructor Documentation

◆ GPUDispatcher()

gem5::GPUDispatcher::GPUDispatcher ( const Params p)

Definition at line 49 of file dispatcher.cc.

References exec().

◆ ~GPUDispatcher()

gem5::GPUDispatcher::~GPUDispatcher ( )

Definition at line 58 of file dispatcher.cc.

Member Function Documentation

◆ dispatch()

void gem5::GPUDispatcher::dispatch ( HSAQueueEntry task)

After all relevant HSA data structures have been traversed/extracted from memory by the CP, dispatch() is called on the dispatcher.

This will schedule a dispatch event that, when triggered, will attempt to dispatch the WGs associated with the given task to the CUs.

Definition at line 114 of file dispatcher.cc.

References gem5::Clocked::clockPeriod(), gem5::curTick(), dispatchActive, gem5::HSAQueueEntry::dispatchId(), DPRINTF, execIds, hsaQueueEntries, gem5::HSAQueueEntry::kernelName(), gem5::GPUDispatcher::GPUDispatcherStats::numKernelLaunched, gem5::EventManager::schedule(), gem5::Event::scheduled(), shader, stats, and tickEvent.

Referenced by gem5::GPUCommandProcessor::dispatchPkt().

◆ exec()

void gem5::GPUDispatcher::exec ( )

There are potentially multiple outstanding kernel launches. It is possible that the workgroups in a different kernel can fit on the GPU even if another kernel's workgroups cannot

dispatch work cannot start until the kernel's invalidate is completely finished; hence, kernel will always initiates invalidate first and keeps waiting until inv done

invalidate is still ongoing, put the kernel on the queue to retry later

if we failed try the next kernel, it may have smaller workgroups. put it on the queue to retry later

Definition at line 133 of file dispatcher.cc.

References gem5::curTick(), gem5::GPUDispatcher::GPUDispatcherStats::cyclesWaitingForDispatch, gem5::Shader::dispatchWorkgroups(), doneIds, DPRINTF, execIds, hsaQueueEntries, gem5::Shader::impl_kern_launch_acq, gem5::Shader::prepareInvalidate(), shader, stats, and gem5::Shader::updateContext().

Referenced by GPUDispatcher().

◆ getOutstandingWbs()

int gem5::GPUDispatcher::getOutstandingWbs ( int  kern_id)

get kernel's outstanding cache writeback requests

Definition at line 280 of file dispatcher.cc.

References hsaQueueEntries.

Referenced by gem5::Shader::prepareFlush().

◆ hsaTask()

HSAQueueEntry * gem5::GPUDispatcher::hsaTask ( int  disp_id)

Definition at line 63 of file dispatcher.cc.

References hsaQueueEntries.

Referenced by gem5::GPUCommandProcessor::submitAgentDispatchPkt().

◆ isReachingKernelEnd()

bool gem5::GPUDispatcher::isReachingKernelEnd ( Wavefront wf)

whether the next workgroup is the final one in the kernel, +1 as we check first before taking action

Definition at line 226 of file dispatcher.cc.

References hsaQueueEntries, and gem5::Wavefront::kernId.

Referenced by gem5::Gcn3ISA::Inst_SOPP__S_ENDPGM::execute(), and gem5::VegaISA::Inst_SOPP__S_ENDPGM::execute().

◆ notifyWgCompl()

void gem5::GPUDispatcher::notifyWgCompl ( Wavefront wf)

When an end program instruction detects that the last WF in a WG has completed it will call this method on the dispatcher.

If we detect that this is the last WG for the given task, then we ring the completion signal, which is used by the CPU to synchronize with the GPU. The HSAPP is also notified that the task has completed so it can be removed from its task queues.

HACK: The semantics of the HSA signal is to decrement the current signal value. We cheat here and read out he value from main memory using functional access and then just DMA the decremented value.

Definition at line 295 of file dispatcher.cc.

References gem5::Clocked::clockPeriod(), gem5::Wavefront::computeUnit, gem5::ComputeUnit::cu_id, gem5::curTick(), DPRINTF, gem5::HSAPacketProcessor::finishPkt(), gem5::GPUCommandProcessor::functionalReadHsaSignal(), gpuCmdProc, gem5::GPUCommandProcessor::hsaPacketProc(), hsaQueueEntries, gem5::Wavefront::kernId, gem5::EventManager::schedule(), gem5::Event::scheduled(), shader, tickEvent, gem5::GPUCommandProcessor::updateHsaSignal(), and gem5::Wavefront::wgId.

Referenced by gem5::Gcn3ISA::Inst_SOPP__S_ENDPGM::execute(), gem5::VegaISA::Inst_SOPP__S_ENDPGM::execute(), and gem5::ComputeUnit::DataPort::handleResponse().

◆ scheduleDispatch()

void gem5::GPUDispatcher::scheduleDispatch ( )

◆ serialize()

void gem5::GPUDispatcher::serialize ( CheckpointOut cp) const
overridevirtual

Serialize an object.

Output an object's state into the current checkpoint section.

Parameters
cpCheckpoint state

Implements gem5::Serializable.

Definition at line 82 of file dispatcher.cc.

References gem5::Event::scheduled(), SERIALIZE_SCALAR, tickEvent, and gem5::Event::when().

◆ setCommandProcessor()

void gem5::GPUDispatcher::setCommandProcessor ( GPUCommandProcessor gpu_cmd_proc)

Definition at line 70 of file dispatcher.cc.

References gpuCmdProc.

Referenced by gem5::GPUCommandProcessor::GPUCommandProcessor().

◆ setShader()

void gem5::GPUDispatcher::setShader ( Shader new_shader)

Definition at line 76 of file dispatcher.cc.

References shader.

◆ unserialize()

void gem5::GPUDispatcher::unserialize ( CheckpointIn cp)
overridevirtual

Unserialize an object.

Read an object's state from the current checkpoint section.

Parameters
cpCheckpoint state

Implements gem5::Serializable.

Definition at line 93 of file dispatcher.cc.

References gem5::EventManager::deschedule(), gem5::EventManager::schedule(), gem5::Event::scheduled(), tickEvent, and UNSERIALIZE_SCALAR.

◆ updateInvCounter()

void gem5::GPUDispatcher::updateInvCounter ( int  kern_id,
int  val = -1 
)

update the counter of oustanding inv requests for the kernel kern_id: kernel id val: +1/-1, increment or decrement the counter (default: -1)

Definition at line 246 of file dispatcher.cc.

References gem5::Clocked::clockPeriod(), gem5::curTick(), hsaQueueEntries, gem5::EventManager::schedule(), gem5::Event::scheduled(), shader, tickEvent, and gem5::X86ISA::val.

Referenced by gem5::ComputeUnit::DataPort::handleResponse(), and gem5::Shader::prepareInvalidate().

◆ updateWbCounter()

bool gem5::GPUDispatcher::updateWbCounter ( int  kern_id,
int  val = -1 
)

update the counter of oustanding wb requests for the kernel kern_id: kernel id val: +1/-1, increment or decrement the counter (default: -1)

return true if all wbs are done for the kernel

Definition at line 266 of file dispatcher.cc.

References hsaQueueEntries, and gem5::X86ISA::val.

Referenced by gem5::ComputeUnit::DataPort::handleResponse(), and gem5::Shader::prepareFlush().

Member Data Documentation

◆ dispatchActive

bool gem5::GPUDispatcher::dispatchActive
private

Definition at line 94 of file dispatcher.hh.

Referenced by dispatch().

◆ doneIds

std::queue<int> gem5::GPUDispatcher::doneIds
private

Definition at line 92 of file dispatcher.hh.

Referenced by exec().

◆ execIds

std::queue<int> gem5::GPUDispatcher::execIds
private

Definition at line 90 of file dispatcher.hh.

Referenced by dispatch(), and exec().

◆ gpuCmdProc

GPUCommandProcessor* gem5::GPUDispatcher::gpuCmdProc
private

Definition at line 86 of file dispatcher.hh.

Referenced by notifyWgCompl(), and setCommandProcessor().

◆ hsaQueueEntries

std::unordered_map<int, HSAQueueEntry*> gem5::GPUDispatcher::hsaQueueEntries
private

◆ shader

Shader* gem5::GPUDispatcher::shader
private

◆ stats

gem5::GPUDispatcher::GPUDispatcherStats gem5::GPUDispatcher::stats
protected

Referenced by dispatch(), and exec().

◆ tickEvent

EventFunctionWrapper gem5::GPUDispatcher::tickEvent
private

The documentation for this class was generated from the following files:

Generated on Wed Dec 21 2022 10:23:23 for gem5 by doxygen 1.9.1