gem5 v24.1.0.1
|
#include <pm4_packet_processor.hh>
Public Member Functions | |
PM4PacketProcessor (const PM4PacketProcessorParams &p) | |
void | setGPUDevice (AMDGPUDevice *gpu_device) |
Tick | write (PacketPtr pkt) override |
Inherited methods. | |
Tick | read (PacketPtr pkt) override |
Pure virtual function that the device must implement. | |
AddrRangeList | getAddrRanges () const override |
Every PIO device is obliged to provide an implementation that returns the address ranges the device responds to. | |
void | serialize (CheckpointOut &cp) const override |
Serialize an object. | |
void | unserialize (CheckpointIn &cp) override |
Unserialize an object. | |
TranslationGenPtr | translate (Addr vaddr, Addr size) override |
Method for functional translation. | |
uint32_t | getKiqDoorbellOffset () |
uint32_t | getPqDoorbellOffset () |
Addr | getGARTAddr (Addr addr) const |
PM4Queue * | getQueue (Addr offset, bool gfx=false) |
Based on an offset communicated through doorbell write, the PM4PacketProcessor identifies which queue needs processing. | |
void | mapPq (Addr offset) |
The first graphics queue, the Primary Queueu a.k.a. | |
void | mapKiq (Addr offset) |
The first compute queue, the Kernel Interface Queueu a.k.a. | |
void | newQueue (QueueDesc *q, Addr offset, PM4MapQueues *pkt=nullptr, int id=-1) |
This method creates a new PM4Queue based on a queue descriptor and an offset. | |
void | process (PM4Queue *q, Addr wptrOffset) |
This method start processing a PM4Queue from the current read pointer to the newly communicated write pointer (i.e., wptrOffset). | |
void | updateReadIndex (Addr offset, uint64_t rd_idx) |
Update read index on doorbell rings. | |
void | decodeNext (PM4Queue *q) |
This method decodes the next packet in a PM4Queue. | |
void | decodeHeader (PM4Queue *q, PM4Header header) |
This method calls other PM4 packet processing methods based on the header of a PM4 packet. | |
void | writeData (PM4Queue *q, PM4WriteData *pkt, PM4Header header) |
void | writeDataDone (PM4Queue *q, PM4WriteData *pkt, Addr addr) |
void | mapQueues (PM4Queue *q, PM4MapQueues *pkt) |
void | unmapQueues (PM4Queue *q, PM4UnmapQueues *pkt) |
void | doneMQDWrite (Addr mqdAddr, Addr addr) |
void | mapProcess (uint32_t pasid, uint64_t ptBase, uint32_t shMemBases) |
void | mapProcessV1 (PM4Queue *q, PM4MapProcess *pkt) |
void | mapProcessV2 (PM4Queue *q, PM4MapProcessV2 *pkt) |
void | processMQD (PM4MapQueues *pkt, PM4Queue *q, Addr addr, QueueDesc *mqd, uint16_t vmid) |
void | processSDMAMQD (PM4MapQueues *pkt, PM4Queue *q, Addr addr, SDMAQueueDesc *mqd, uint16_t vmid) |
void | releaseMem (PM4Queue *q, PM4ReleaseMem *pkt) |
void | releaseMemDone (PM4Queue *q, PM4ReleaseMem *pkt, Addr addr) |
void | runList (PM4Queue *q, PM4RunList *pkt) |
void | indirectBuffer (PM4Queue *q, PM4IndirectBuf *pkt) |
void | switchBuffer (PM4Queue *q, PM4SwitchBuf *pkt) |
void | setUconfigReg (PM4Queue *q, PM4SetUconfigReg *pkt) |
void | waitRegMem (PM4Queue *q, PM4WaitRegMem *pkt) |
void | queryStatus (PM4Queue *q, PM4QueryStatus *pkt) |
void | queryStatusDone (PM4Queue *q, PM4QueryStatus *pkt) |
void | writeMMIO (PacketPtr pkt, Addr mmio_offset) |
void | setHqdVmid (uint32_t data) |
void | setHqdActive (uint32_t data) |
void | setHqdPqBase (uint32_t data) |
void | setHqdPqBaseHi (uint32_t data) |
void | setHqdPqDoorbellCtrl (uint32_t data) |
void | setHqdPqPtr (uint32_t data) |
void | setHqdPqWptrLo (uint32_t data) |
void | setHqdPqWptrHi (uint32_t data) |
void | setHqdPqRptrReportAddr (uint32_t data) |
void | setHqdPqRptrReportAddrHi (uint32_t data) |
void | setHqdPqWptrPollAddr (uint32_t data) |
void | setHqdPqWptrPollAddrHi (uint32_t data) |
void | setHqdPqControl (uint32_t data) |
void | setHqdIbCtrl (uint32_t data) |
void | setRbVmid (uint32_t data) |
void | setRbCntl (uint32_t data) |
void | setRbWptrLo (uint32_t data) |
void | setRbWptrHi (uint32_t data) |
void | setRbRptrAddrLo (uint32_t data) |
void | setRbRptrAddrHi (uint32_t data) |
void | setRbWptrPollAddrLo (uint32_t data) |
void | setRbWptrPollAddrHi (uint32_t data) |
void | setRbBaseLo (uint32_t data) |
void | setRbBaseHi (uint32_t data) |
void | setRbDoorbellCntrl (uint32_t data) |
void | setRbDoorbellRangeLo (uint32_t data) |
void | setRbDoorbellRangeHi (uint32_t data) |
int | getIpId () const |
AddrRange | getMMIORange () const |
![]() | |
DmaVirtDevice (const Params &p) | |
virtual | ~DmaVirtDevice () |
void | dmaReadVirt (Addr host_addr, unsigned size, DmaCallback *cb, void *data, Tick delay=0) |
Initiate a DMA read from virtual address host_addr. | |
void | dmaWriteVirt (Addr host_addr, unsigned size, DmaCallback *b, void *data, Tick delay=0) |
Initiate a DMA write from virtual address host_addr. | |
void | dmaVirt (DmaFnPtr dmaFn, Addr host_addr, unsigned size, DmaCallback *cb, void *data, Tick delay=0) |
Initiate a call to DmaDevice using DmaFnPtr do a DMA starting from virtual address host_addr for size number of bytes on the data. | |
![]() | |
DmaDevice (const Params &p) | |
virtual | ~DmaDevice ()=default |
void | dmaWrite (Addr addr, int size, Event *event, uint8_t *data, uint32_t sid, uint32_t ssid, Tick delay=0) |
void | dmaWrite (Addr addr, int size, Event *event, uint8_t *data, Tick delay=0) |
void | dmaRead (Addr addr, int size, Event *event, uint8_t *data, uint32_t sid, uint32_t ssid, Tick delay=0) |
void | dmaRead (Addr addr, int size, Event *event, uint8_t *data, Tick delay=0) |
bool | dmaPending () const |
void | init () override |
init() is called after all C++ SimObjects have been created and all ports are connected. | |
Addr | cacheBlockSize () const |
Port & | getPort (const std::string &if_name, PortID idx=InvalidPortID) override |
Get a port with a given name and index. | |
![]() | |
PioDevice (const Params &p) | |
virtual | ~PioDevice () |
void | init () override |
init() is called after all C++ SimObjects have been created and all ports are connected. | |
Port & | getPort (const std::string &if_name, PortID idx=InvalidPortID) override |
Get a port with a given name and index. | |
![]() | |
ClockedObject (const ClockedObjectParams &p) | |
![]() | |
const Params & | params () const |
SimObject (const Params &p) | |
virtual | ~SimObject () |
virtual void | loadState (CheckpointIn &cp) |
loadState() is called on each SimObject when restoring from a checkpoint. | |
virtual void | initState () |
initState() is called on each SimObject when not restoring from a checkpoint. | |
virtual void | regProbePoints () |
Register probe points for this object. | |
virtual void | regProbeListeners () |
Register probe listeners for this object. | |
ProbeManager * | getProbeManager () |
Get the probe manager for this object. | |
virtual void | startup () |
startup() is the final initialization call before simulation. | |
DrainState | drain () override |
Provide a default implementation of the drain interface for objects that don't need draining. | |
virtual void | memWriteback () |
Write back dirty buffers to memory using functional writes. | |
virtual void | memInvalidate () |
Invalidate the contents of memory buffers. | |
void | serialize (CheckpointOut &cp) const override |
Serialize an object. | |
void | unserialize (CheckpointIn &cp) override |
Unserialize an object. | |
![]() | |
EventQueue * | eventQueue () 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. | |
void | setCurTick (Tick newVal) |
EventManager (EventManager &em) | |
Event manger manages events in the event queue. | |
EventManager (EventManager *em) | |
EventManager (EventQueue *eq) | |
![]() | |
Serializable () | |
virtual | ~Serializable () |
void | serializeSection (CheckpointOut &cp, const char *name) const |
Serialize an object into a new section. | |
void | serializeSection (CheckpointOut &cp, const std::string &name) const |
void | unserializeSection (CheckpointIn &cp, const char *name) |
Unserialize an a child object. | |
void | unserializeSection (CheckpointIn &cp, const std::string &name) |
![]() | |
DrainState | drainState () const |
Return the current drain state of an object. | |
virtual void | notifyFork () |
Notify a child process of a fork. | |
![]() | |
Group (Group *parent, const char *name=nullptr) | |
Construct a new statistics group. | |
virtual | ~Group () |
virtual void | regStats () |
Callback to set stat parameters. | |
virtual void | resetStats () |
Callback to reset stats. | |
virtual void | preDumpStats () |
Callback before stats are dumped. | |
void | addStat (statistics::Info *info) |
Register a stat with this group. | |
const std::map< std::string, Group * > & | getStatGroups () const |
Get all child groups associated with this object. | |
const std::vector< Info * > & | getStats () const |
Get all stats associated with this object. | |
void | addStatGroup (const char *name, Group *block) |
Add a stat block as a child of this block. | |
const Info * | resolveStat (std::string name) const |
Resolve a stat by its name within this group. | |
void | mergeStatGroup (Group *block) |
Merge the contents (stats & children) of a block to this block. | |
Group (const Group &)=delete | |
Group & | operator= (const Group &)=delete |
![]() | |
Named (const std::string &name_) | |
virtual | ~Named ()=default |
virtual std::string | name () const |
![]() | |
void | updateClockPeriod () |
Update the tick to the current tick. | |
Tick | clockEdge (Cycles cycles=Cycles(0)) const |
Determine the tick when a cycle begins, by default the current one, but the argument also enables the caller to determine a future cycle. | |
Cycles | curCycle () const |
Determine the current cycle, corresponding to a tick aligned to a clock edge. | |
Tick | nextCycle () const |
Based on the clock of the object, determine the start tick of the first cycle that is at least one cycle in the future. | |
uint64_t | frequency () const |
Tick | clockPeriod () const |
double | voltage () const |
Cycles | ticksToCycles (Tick t) const |
Tick | cyclesToTicks (Cycles c) const |
Private Member Functions | |
void | unmapAllQueues (bool unmap_static) |
Private Attributes | |
AMDGPUDevice * | gpuDevice |
PrimaryQueue | pq |
PM4MapQueues | pq_pkt |
QueueDesc | kiq |
PM4MapQueues | kiq_pkt |
std::unordered_map< uint16_t, PM4Queue * > | queues |
std::unordered_map< uint32_t, PM4Queue * > | queuesMap |
int | _ipId |
AddrRange | _mmioRange |
Additional Inherited Members | |
![]() | |
typedef void(DmaDevice::* | DmaFnPtr) (Addr, int, Event *, uint8_t *, Tick) |
![]() | |
typedef DmaDeviceParams | Params |
![]() | |
using | Params = PioDeviceParams |
![]() | |
using | Params = ClockedObjectParams |
Parameters of ClockedObject. | |
![]() | |
typedef SimObjectParams | Params |
![]() | |
static void | serializeAll (const std::string &cpt_dir) |
Create a checkpoint by serializing all SimObjects in the system. | |
static SimObject * | find (const char *name) |
Find the SimObject with the given name and return a pointer to it. | |
static void | setSimObjectResolver (SimObjectResolver *resolver) |
There is a single object name resolver, and it is only set when simulation is restoring from checkpoints. | |
static SimObjectResolver * | getSimObjectResolver () |
There is a single object name resolver, and it is only set when simulation is restoring from checkpoints. | |
![]() | |
static const std::string & | currentSection () |
Gets the fully-qualified name of the active section. | |
static void | generateCheckpointOut (const std::string &cpt_dir, std::ofstream &outstream) |
Generate a checkpoint file so that the serialization can be routed to it. | |
![]() | |
PowerState * | powerState |
![]() | |
Drainable () | |
virtual | ~Drainable () |
virtual void | drainResume () |
Resume execution after a successful drain. | |
void | signalDrainDone () const |
Signal that an object is drained. | |
![]() | |
Clocked (ClockDomain &clk_domain) | |
Create a clocked object and set the clock domain based on the parameters. | |
Clocked (Clocked &)=delete | |
Clocked & | operator= (Clocked &)=delete |
virtual | ~Clocked () |
Virtual destructor due to inheritance. | |
void | resetClock () const |
Reset the object's clock using the current global tick value. | |
virtual void | clockPeriodUpdated () |
A hook subclasses can implement so they can do any extra work that's needed when the clock rate is changed. | |
![]() | |
DmaPort | dmaPort |
![]() | |
System * | sys |
PioPort< PioDevice > | pioPort |
The pioPort that handles the requests for us and provides us requests that it sees. | |
![]() | |
const SimObjectParams & | _params |
Cached copy of the object parameters. | |
![]() | |
EventQueue * | eventq |
A pointer to this object's event queue. | |
Definition at line 52 of file pm4_packet_processor.hh.
gem5::PM4PacketProcessor::PM4PacketProcessor | ( | const PM4PacketProcessorParams & | p | ) |
Definition at line 51 of file pm4_packet_processor.cc.
This method calls other PM4 packet processing methods based on the header of a PM4 packet.
Definition at line 210 of file pm4_packet_processor.cc.
References decodeNext(), gem5::DmaVirtDevice::dmaReadVirt(), DPRINTF, getGARTAddr(), gem5::AMDGPUDevice::getGfxVersion(), gem5::AMDGPUDevice::getVM(), gpuDevice, header, indirectBuffer(), gem5::AMDGPUVM::invalidateTLBs(), gem5::IT_INDIRECT_BUFFER, gem5::IT_INVALIDATE_TLBS, gem5::IT_MAP_PROCESS, gem5::IT_MAP_QUEUES, gem5::IT_NOP, gem5::IT_QUERY_STATUS, gem5::IT_RELEASE_MEM, gem5::IT_RUN_LIST, gem5::IT_SET_UCONFIG_REG, gem5::IT_SWITCH_BUFFER, gem5::IT_UNMAP_QUEUES, gem5::IT_WAIT_REG_MEM, gem5::IT_WRITE_DATA, mapProcessV1(), mapProcessV2(), mapQueues(), gem5::ArmISA::q, queryStatus(), releaseMem(), runList(), setUconfigReg(), switchBuffer(), unmapQueues(), waitRegMem(), warn, and writeData().
Referenced by decodeNext().
void gem5::PM4PacketProcessor::decodeNext | ( | PM4Queue * | q | ) |
This method decodes the next packet in a PM4Queue.
Definition at line 166 of file pm4_packet_processor.cc.
References gem5::X86ISA::addr, data, decodeHeader(), decodeNext(), gem5::DmaVirtDevice::dmaReadVirt(), gem5::DmaVirtDevice::dmaWriteVirt(), DPRINTF, getGARTAddr(), header, and gem5::ArmISA::q.
Referenced by decodeHeader(), decodeNext(), indirectBuffer(), mapProcessV1(), mapProcessV2(), process(), processMQD(), processSDMAMQD(), queryStatusDone(), releaseMemDone(), runList(), setUconfigReg(), switchBuffer(), unmapQueues(), waitRegMem(), writeData(), and writeDataDone().
Definition at line 698 of file pm4_packet_processor.cc.
References gem5::X86ISA::addr, and DPRINTF.
Referenced by unmapAllQueues().
|
overridevirtual |
Every PIO device is obliged to provide an implementation that returns the address ranges the device responds to.
Implements gem5::PioDevice.
Definition at line 79 of file pm4_packet_processor.cc.
Definition at line 92 of file pm4_packet_processor.cc.
References gem5::X86ISA::addr, gem5::bits(), gem5::AMDGPUDevice::getVM(), gpuDevice, and gem5::AMDGPUVM::inAGP().
Referenced by decodeHeader(), decodeNext(), mapQueues(), queryStatus(), releaseMem(), unmapAllQueues(), and writeData().
|
inline |
Definition at line 198 of file pm4_packet_processor.hh.
References _ipId.
Referenced by newQueue(), processSDMAMQD(), setUconfigReg(), and writeMMIO().
|
inline |
Definition at line 91 of file pm4_packet_processor.hh.
References gem5::GEM5_PACKED::doorbell, and kiq.
Referenced by writeMMIO().
|
inline |
Definition at line 199 of file pm4_packet_processor.hh.
References _mmioRange.
|
inline |
Definition at line 92 of file pm4_packet_processor.hh.
References gem5::PrimaryQueue::doorbellOffset, and pq.
Referenced by writeMMIO().
Based on an offset communicated through doorbell write, the PM4PacketProcessor identifies which queue needs processing.
Definition at line 102 of file pm4_packet_processor.cc.
References mapKiq(), mapPq(), gem5::ArmISA::offset, and queuesMap.
void gem5::PM4PacketProcessor::indirectBuffer | ( | PM4Queue * | q, |
PM4IndirectBuf * | pkt | ||
) |
Definition at line 771 of file pm4_packet_processor.cc.
References decodeNext(), DPRINTF, gem5::GEM5_PACKED::ibBase, gem5::GEM5_PACKED::ibSize, and gem5::ArmISA::q.
Referenced by decodeHeader().
void gem5::PM4PacketProcessor::mapKiq | ( | Addr | offset | ) |
The first compute queue, the Kernel Interface Queueu a.k.a.
KIQ, needs to be mapped since all queue details are communicated through MMIOs to special registers.
Definition at line 116 of file pm4_packet_processor.cc.
References DPRINTF, kiq, kiq_pkt, newQueue(), and gem5::ArmISA::offset.
Referenced by getQueue().
void gem5::PM4PacketProcessor::mapPq | ( | Addr | offset | ) |
The first graphics queue, the Primary Queueu a.k.a.
RB0, needs to be mapped since all queue details are communicated through MMIOs to special registers.
Definition at line 123 of file pm4_packet_processor.cc.
References DPRINTF, newQueue(), gem5::ArmISA::offset, pq, and pq_pkt.
Referenced by getQueue().
void gem5::PM4PacketProcessor::mapProcess | ( | uint32_t | pasid, |
uint64_t | ptBase, | ||
uint32_t | shMemBases | ||
) |
Definition at line 704 of file pm4_packet_processor.cc.
References gem5::AMDGPUDevice::allocateVMID(), gem5::bits(), gem5::AMDGPUDevice::CP(), gem5::AMDGPUDevice::getVM(), gpuDevice, gem5::HW_REG_SH_MEM_BASES, gem5::pasid, gem5::Shader::setHwReg(), gem5::Shader::setLdsApe(), gem5::AMDGPUVM::setPageTableBase(), gem5::Shader::setScratchApe(), and gem5::GPUCommandProcessor::shader().
Referenced by mapProcessV1(), and mapProcessV2().
void gem5::PM4PacketProcessor::mapProcessV1 | ( | PM4Queue * | q, |
PM4MapProcess * | pkt | ||
) |
Definition at line 724 of file pm4_packet_processor.cc.
References gem5::GEM5_PACKED::completionSignal, decodeNext(), DPRINTF, mapProcess(), gem5::GEM5_PACKED::pasid, gem5::GEM5_PACKED::processQuantum, gem5::GEM5_PACKED::ptBase, gem5::ArmISA::q, and gem5::GEM5_PACKED::shMemBases.
Referenced by decodeHeader().
void gem5::PM4PacketProcessor::mapProcessV2 | ( | PM4Queue * | q, |
PM4MapProcessV2 * | pkt | ||
) |
Definition at line 739 of file pm4_packet_processor.cc.
References gem5::GEM5_PACKED::completionSignal, decodeNext(), DPRINTF, mapProcess(), gem5::GEM5_PACKED::pasid, gem5::GEM5_PACKED::processQuantum, gem5::GEM5_PACKED::ptBase, gem5::ArmISA::q, and gem5::GEM5_PACKED::shMemBases.
Referenced by decodeHeader().
void gem5::PM4PacketProcessor::mapQueues | ( | PM4Queue * | q, |
PM4MapQueues * | pkt | ||
) |
Definition at line 412 of file pm4_packet_processor.cc.
References gem5::X86ISA::addr, gem5::GEM5_PACKED::allocFormat, gem5::GEM5_PACKED::checkDisable, gem5::DmaVirtDevice::dmaReadVirt(), gem5::GEM5_PACKED::doorbellOffset, DPRINTF, gem5::GEM5_PACKED::engineSel, getGARTAddr(), gpuDevice, gem5::AMDGPUDevice::lastVMID(), gem5::AMDGPUDevice::mapDoorbellToVMID(), gem5::GEM5_PACKED::me, gem5::GEM5_PACKED::mqdAddr, gem5::GEM5_PACKED::numQueues, panic, gem5::GEM5_PACKED::pipe, processMQD(), processSDMAMQD(), gem5::ArmISA::q, gem5::GEM5_PACKED::queueSel, gem5::GEM5_PACKED::queueSlot, gem5::GEM5_PACKED::queueType, gem5::GEM5_PACKED::vmid, and gem5::GEM5_PACKED::wptrAddr.
Referenced by decodeHeader().
void gem5::PM4PacketProcessor::newQueue | ( | QueueDesc * | q, |
Addr | offset, | ||
PM4MapQueues * | pkt = nullptr , |
||
int | id = -1 |
||
) |
This method creates a new PM4Queue based on a queue descriptor and an offset.
Definition at line 130 of file pm4_packet_processor.cc.
References gem5::GEM5_PACKED::aql, gem5::GEM5_PACKED::base, gem5::Compute, gem5::ComputeAQL, DPRINTF, getIpId(), gpuDevice, gem5::ArmISA::id, gem5::ArmISA::offset, gem5::ArmISA::q, queues, queuesMap, and gem5::AMDGPUDevice::setDoorbellType().
Referenced by mapKiq(), mapPq(), processMQD(), and unserialize().
This method start processing a PM4Queue from the current read pointer to the newly communicated write pointer (i.e., wptrOffset).
Definition at line 155 of file pm4_packet_processor.cc.
References decodeNext(), and gem5::ArmISA::q.
void gem5::PM4PacketProcessor::processMQD | ( | PM4MapQueues * | pkt, |
PM4Queue * | q, | ||
Addr | addr, | ||
QueueDesc * | mqd, | ||
uint16_t | vmid | ||
) |
Definition at line 463 of file pm4_packet_processor.cc.
References gem5::GEM5_PACKED::aql, gem5::GEM5_PACKED::aqlRptr, gem5::GEM5_PACKED::base, gem5::AMDGPUDevice::CP(), decodeNext(), gem5::GEM5_PACKED::doorbell, DPRINTF, gpuDevice, gem5::GEM5_PACKED::hqd_active, gem5::GEM5_PACKED::hqd_pq_control, gem5::GEM5_PACKED::hqd_vmid, gem5::GPUCommandProcessor::hsaPacketProc(), gem5::PM4Queue::id(), gem5::AMDGPUDevice::insertQId(), gem5::GEM5_PACKED::mqdBase, gem5::GEM5_PACKED::mqdReadIndex, newQueue(), gem5::ArmISA::offset, gem5::AMDGPUDevice::processPendingDoorbells(), gem5::ArmISA::q, queuesMap, gem5::GEM5_PACKED::rptr, and gem5::HSAPacketProcessor::setDeviceQueueDesc().
Referenced by mapQueues().
void gem5::PM4PacketProcessor::processSDMAMQD | ( | PM4MapQueues * | pkt, |
PM4Queue * | q, | ||
Addr | addr, | ||
SDMAQueueDesc * | mqd, | ||
uint16_t | vmid | ||
) |
Definition at line 502 of file pm4_packet_processor.cc.
References gem5::X86ISA::addr, gem5::bits(), decodeNext(), gem5::GEM5_PACKED::doorbellOffset, DPRINTF, gem5::GEM5_PACKED::engineSel, getIpId(), gem5::AMDGPUDevice::getSDMAById(), gpuDevice, gem5::AMDGPUDevice::processPendingDoorbells(), gem5::ArmISA::q, gem5::GEM5_PACKED::queueType, gem5::GEM5_PACKED::rb_base, gem5::SDMAEngine::registerRLCQueue(), gem5::RLC, gem5::GEM5_PACKED::sdmax_rlcx_ib_base_hi, gem5::GEM5_PACKED::sdmax_rlcx_ib_base_lo, gem5::GEM5_PACKED::sdmax_rlcx_rb_cntl, gem5::GEM5_PACKED::sdmax_rlcx_rb_rptr, gem5::GEM5_PACKED::sdmax_rlcx_rb_rptr_addr_hi, gem5::GEM5_PACKED::sdmax_rlcx_rb_rptr_addr_lo, gem5::GEM5_PACKED::sdmax_rlcx_rb_rptr_hi, gem5::GEM5_PACKED::sdmax_rlcx_rb_wptr, gem5::GEM5_PACKED::sdmax_rlcx_rb_wptr_hi, gem5::AMDGPUDevice::setDoorbellType(), and gem5::AMDGPUDevice::setSDMAEngine().
Referenced by mapQueues().
void gem5::PM4PacketProcessor::queryStatus | ( | PM4Queue * | q, |
PM4QueryStatus * | pkt | ||
) |
Definition at line 836 of file pm4_packet_processor.cc.
References gem5::X86ISA::addr, gem5::GEM5_PACKED::addr, gem5::GEM5_PACKED::command, gem5::GEM5_PACKED::contextId, gem5::GEM5_PACKED::data, gem5::DmaVirtDevice::dmaWriteVirt(), gem5::GEM5_PACKED::doorbellOffset, DPRINTF, gem5::GEM5_PACKED::engineSel, getGARTAddr(), gem5::GEM5_PACKED::interruptSel, panic, gem5::GEM5_PACKED::pasid, gem5::ArmISA::q, and queryStatusDone().
Referenced by decodeHeader().
void gem5::PM4PacketProcessor::queryStatusDone | ( | PM4Queue * | q, |
PM4QueryStatus * | pkt | ||
) |
Definition at line 861 of file pm4_packet_processor.cc.
References decodeNext(), DPRINTF, and gem5::ArmISA::q.
Referenced by queryStatus().
Pure virtual function that the device must implement.
Called when a read command is recieved by the port.
pkt | Packet describing this request |
Implements gem5::PioDevice.
Definition at line 81 of file pm4_packet_processor.hh.
void gem5::PM4PacketProcessor::releaseMem | ( | PM4Queue * | q, |
PM4ReleaseMem * | pkt | ||
) |
Definition at line 538 of file pm4_packet_processor.cc.
References gem5::X86ISA::addr, gem5::GEM5_PACKED::addr, gem5::GEM5_PACKED::dataLo, gem5::GEM5_PACKED::dataSelect, gem5::GEM5_PACKED::destSelect, gem5::DmaVirtDevice::dmaWriteVirt(), DPRINTF, gem5::GEM5_PACKED::event, gem5::GEM5_PACKED::eventIdx, getGARTAddr(), gem5::GEM5_PACKED::intCtxId, gem5::GEM5_PACKED::intSelect, panic, gem5::ArmISA::q, and releaseMemDone().
Referenced by decodeHeader().
void gem5::PM4PacketProcessor::releaseMemDone | ( | PM4Queue * | q, |
PM4ReleaseMem * | pkt, | ||
Addr | addr | ||
) |
Definition at line 562 of file pm4_packet_processor.cc.
References gem5::X86ISA::addr, gem5::CP_EOP, gem5::GEM5_PACKED::dataLo, decodeNext(), DPRINTF, gem5::AMDGPUDevice::getIH(), gpuDevice, gem5::GEM5_PACKED::intCtxId, gem5::GEM5_PACKED::intSelect, gem5::AMDGPUInterruptHandler::prepareInterruptCookie(), gem5::ArmISA::q, gem5::SOC15_IH_CLIENTID_GRBM_CP, and gem5::AMDGPUInterruptHandler::submitInterruptCookie().
Referenced by releaseMem().
void gem5::PM4PacketProcessor::runList | ( | PM4Queue * | q, |
PM4RunList * | pkt | ||
) |
Definition at line 754 of file pm4_packet_processor.cc.
References decodeNext(), DPRINTF, gem5::GEM5_PACKED::ibBase, gem5::GEM5_PACKED::ibSize, and gem5::ArmISA::q.
Referenced by decodeHeader().
|
overridevirtual |
Serialize an object.
Output an object's state into the current checkpoint section.
cp | Checkpoint state |
Reimplemented from gem5::ClockedObject.
Definition at line 1127 of file pm4_packet_processor.cc.
References gem5::X86ISA::base, gem5::ArmISA::i, gem5::PM4Queue::id(), gem5::PowerISA::me, gem5::ArmISA::offset, gem5::ArmISA::q, queues, gem5::ClockedObject::serialize(), SERIALIZE_ARRAY, and SERIALIZE_SCALAR.
void gem5::PM4PacketProcessor::setGPUDevice | ( | AMDGPUDevice * | gpu_device | ) |
Definition at line 86 of file pm4_packet_processor.cc.
References gpuDevice.
void gem5::PM4PacketProcessor::setHqdActive | ( | uint32_t | data | ) |
Definition at line 970 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_active, and kiq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setHqdIbCtrl | ( | uint32_t | data | ) |
Definition at line 1042 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_ib_control, and kiq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setHqdPqBase | ( | uint32_t | data | ) |
Definition at line 976 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_pq_base_lo, and kiq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setHqdPqBaseHi | ( | uint32_t | data | ) |
Definition at line 982 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_pq_base_hi, and kiq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setHqdPqControl | ( | uint32_t | data | ) |
Definition at line 1036 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_pq_control, and kiq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setHqdPqDoorbellCtrl | ( | uint32_t | data | ) |
Definition at line 988 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_pq_doorbell_control, and kiq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setHqdPqPtr | ( | uint32_t | data | ) |
Definition at line 994 of file pm4_packet_processor.cc.
References data, kiq, and gem5::GEM5_PACKED::rptr.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setHqdPqRptrReportAddr | ( | uint32_t | data | ) |
Definition at line 1012 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_pq_rptr_report_addr_lo, and kiq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setHqdPqRptrReportAddrHi | ( | uint32_t | data | ) |
Definition at line 1018 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_pq_rptr_report_addr_hi, and kiq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setHqdPqWptrHi | ( | uint32_t | data | ) |
Definition at line 1006 of file pm4_packet_processor.cc.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setHqdPqWptrLo | ( | uint32_t | data | ) |
Definition at line 1000 of file pm4_packet_processor.cc.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setHqdPqWptrPollAddr | ( | uint32_t | data | ) |
Definition at line 1024 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_pq_wptr_poll_addr_lo, and kiq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setHqdPqWptrPollAddrHi | ( | uint32_t | data | ) |
Definition at line 1030 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_pq_wptr_poll_addr_hi, and kiq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setHqdVmid | ( | uint32_t | data | ) |
Definition at line 964 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_vmid, and kiq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setRbBaseHi | ( | uint32_t | data | ) |
Definition at line 1102 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_pq_base_hi, and pq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setRbBaseLo | ( | uint32_t | data | ) |
Definition at line 1096 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_pq_base_lo, and pq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setRbCntl | ( | uint32_t | data | ) |
Definition at line 1054 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_pq_control, and pq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setRbDoorbellCntrl | ( | uint32_t | data | ) |
Definition at line 1108 of file pm4_packet_processor.cc.
References data, gem5::PrimaryQueue::doorbellOffset, gem5::GEM5_PACKED::hqd_pq_doorbell_control, and pq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setRbDoorbellRangeHi | ( | uint32_t | data | ) |
Definition at line 1121 of file pm4_packet_processor.cc.
References data, gem5::PrimaryQueue::doorbellRangeHi, and pq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setRbDoorbellRangeLo | ( | uint32_t | data | ) |
Definition at line 1115 of file pm4_packet_processor.cc.
References data, gem5::PrimaryQueue::doorbellRangeLo, and pq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setRbRptrAddrHi | ( | uint32_t | data | ) |
Definition at line 1078 of file pm4_packet_processor.cc.
References data, pq, and gem5::PrimaryQueue::queueRptrAddrHi.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setRbRptrAddrLo | ( | uint32_t | data | ) |
Definition at line 1072 of file pm4_packet_processor.cc.
References data, pq, and gem5::PrimaryQueue::queueRptrAddrLo.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setRbVmid | ( | uint32_t | data | ) |
Definition at line 1048 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_vmid, and pq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setRbWptrHi | ( | uint32_t | data | ) |
Definition at line 1066 of file pm4_packet_processor.cc.
References data, pq, and gem5::PrimaryQueue::queueWptrHi.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setRbWptrLo | ( | uint32_t | data | ) |
Definition at line 1060 of file pm4_packet_processor.cc.
References data, pq, and gem5::PrimaryQueue::queueWptrLo.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setRbWptrPollAddrHi | ( | uint32_t | data | ) |
Definition at line 1090 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_pq_wptr_poll_addr_hi, and pq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setRbWptrPollAddrLo | ( | uint32_t | data | ) |
Definition at line 1084 of file pm4_packet_processor.cc.
References data, gem5::GEM5_PACKED::hqd_pq_wptr_poll_addr_lo, and pq.
Referenced by writeMMIO().
void gem5::PM4PacketProcessor::setUconfigReg | ( | PM4Queue * | q, |
PM4SetUconfigReg * | pkt | ||
) |
Definition at line 800 of file pm4_packet_processor.cc.
References gem5::GEM5_PACKED::data, decodeNext(), DPRINTF, getIpId(), gpuDevice, gem5::GEM5_PACKED::offset, PACKET3_SET_UCONFIG_REG_START, gem5::ArmISA::q, and gem5::AMDGPUDevice::setRegVal().
Referenced by decodeHeader().
void gem5::PM4PacketProcessor::switchBuffer | ( | PM4Queue * | q, |
PM4SwitchBuf * | pkt | ||
) |
Definition at line 787 of file pm4_packet_processor.cc.
References decodeNext(), DPRINTF, and gem5::ArmISA::q.
Referenced by decodeHeader().
|
overridevirtual |
Method for functional translation.
AMDGPUDevice will perform DMA operations on VAs, and because page faults are not currently supported for Vega 10, we must be able to find the pages mapped for the process.
Implements gem5::DmaVirtDevice.
Definition at line 64 of file pm4_packet_processor.cc.
References gem5::AMDGPUDevice::getVM(), gpuDevice, gem5::AMDGPUVM::inAGP(), and gem5::MipsISA::vaddr.
|
private |
Definition at line 593 of file pm4_packet_processor.cc.
References gem5::X86ISA::addr, gem5::GEM5_PACKED::base, gem5::AMDGPUDevice::CP(), gem5::DmaVirtDevice::dmaWriteVirt(), doneMQDWrite(), DPRINTF, getGARTAddr(), gem5::AMDGPUDevice::getUsedVMIDs(), gpuDevice, gem5::GPUCommandProcessor::hsaPacketProc(), gem5::ArmISA::id, gem5::GEM5_PACKED::mqdReadIndex, and queues.
Referenced by unmapQueues().
void gem5::PM4PacketProcessor::unmapQueues | ( | PM4Queue * | q, |
PM4UnmapQueues * | pkt | ||
) |
Definition at line 634 of file pm4_packet_processor.cc.
References gem5::AMDGPUDevice::deallocateAllQueues(), gem5::AMDGPUDevice::deallocatePasid(), gem5::AMDGPUDevice::deallocateVmid(), decodeNext(), gem5::GEM5_PACKED::doorbellOffset0, gem5::GEM5_PACKED::doorbellOffset1, gem5::GEM5_PACKED::doorbellOffset2, gem5::GEM5_PACKED::doorbellOffset3, DPRINTF, gem5::AMDGPUDevice::getVMID(), gpuDevice, gem5::GEM5_PACKED::numQueues, panic, gem5::GEM5_PACKED::pasid, gem5::ArmISA::q, gem5::GEM5_PACKED::queueSel, and unmapAllQueues().
Referenced by decodeHeader().
|
overridevirtual |
Unserialize an object.
Read an object's state from the current checkpoint section.
cp | Checkpoint state |
Reimplemented from gem5::ClockedObject.
Definition at line 1217 of file pm4_packet_processor.cc.
References gem5::GEM5_PACKED::aql, gem5::X86ISA::base, gem5::GEM5_PACKED::base, gem5::AMDGPUDevice::CP(), DPRINTF, gpuDevice, gem5::GPUCommandProcessor::hsaPacketProc(), gem5::ArmISA::i, gem5::PowerISA::me, gem5::GEM5_PACKED::mqdBase, gem5::GEM5_PACKED::mqdReadIndex, newQueue(), gem5::ArmISA::offset, queues, gem5::HSAPacketProcessor::setDeviceQueueDesc(), gem5::ClockedObject::unserialize(), UNSERIALIZE_ARRAY, and UNSERIALIZE_SCALAR.
void gem5::PM4PacketProcessor::updateReadIndex | ( | Addr | offset, |
uint64_t | rd_idx | ||
) |
Update read index on doorbell rings.
We use write index, however read index == write index when the queue is empty. This allows us to save previous read index when a queue is remapped. The remapped queue will read from the previous read index rather than reset to zero.
Definition at line 586 of file pm4_packet_processor.cc.
References gem5::ArmISA::offset, and queuesMap.
void gem5::PM4PacketProcessor::waitRegMem | ( | PM4Queue * | q, |
PM4WaitRegMem * | pkt | ||
) |
Definition at line 819 of file pm4_packet_processor.cc.
References decodeNext(), DPRINTF, gem5::GEM5_PACKED::function, gem5::GEM5_PACKED::mask, gem5::GEM5_PACKED::memAddrHi, gem5::GEM5_PACKED::memAddrLo, gem5::GEM5_PACKED::memSpace, gem5::GEM5_PACKED::operation, gem5::GEM5_PACKED::pollInterval, gem5::ArmISA::q, and gem5::GEM5_PACKED::reference.
Referenced by decodeHeader().
Inherited methods.
Implements gem5::PioDevice.
Definition at line 80 of file pm4_packet_processor.hh.
void gem5::PM4PacketProcessor::writeData | ( | PM4Queue * | q, |
PM4WriteData * | pkt, | ||
PM4Header | header | ||
) |
Definition at line 356 of file pm4_packet_processor.cc.
References gem5::X86ISA::addr, gem5::GEM5_PACKED::addrIncr, gem5::GEM5_PACKED::cachePolicy, gem5::GEM5_PACKED::data, decodeNext(), gem5::GEM5_PACKED::destAddr, gem5::GEM5_PACKED::destSel, gem5::DmaVirtDevice::dmaWriteVirt(), DPRINTF, fatal, getGARTAddr(), gpuDevice, header, gem5::ArmISA::q, gem5::GEM5_PACKED::resume, gem5::AMDGPUDevice::setRegVal(), gem5::GEM5_PACKED::writeConfirm, and writeDataDone().
Referenced by decodeHeader().
void gem5::PM4PacketProcessor::writeDataDone | ( | PM4Queue * | q, |
PM4WriteData * | pkt, | ||
Addr | addr | ||
) |
Definition at line 399 of file pm4_packet_processor.cc.
References gem5::X86ISA::addr, gem5::GEM5_PACKED::data, decodeNext(), DPRINTF, gem5::ArmISA::q, and gem5::GEM5_PACKED::writeConfirm.
Referenced by writeData().
Definition at line 870 of file pm4_packet_processor.cc.
References gem5::Compute, getIpId(), getKiqDoorbellOffset(), gem5::Packet::getLE(), getPqDoorbellOffset(), gem5::Gfx, gpuDevice, mmCP_HQD_ACTIVE, mmCP_HQD_IB_CONTROL, mmCP_HQD_PQ_BASE, mmCP_HQD_PQ_BASE_HI, mmCP_HQD_PQ_CONTROL, mmCP_HQD_PQ_DOORBELL_CONTROL, mmCP_HQD_PQ_RPTR, mmCP_HQD_PQ_RPTR_REPORT_ADDR, mmCP_HQD_PQ_RPTR_REPORT_ADDR_HI, mmCP_HQD_PQ_WPTR_HI, mmCP_HQD_PQ_WPTR_LO, mmCP_HQD_PQ_WPTR_POLL_ADDR, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI, mmCP_HQD_VMID, mmCP_RB0_BASE, mmCP_RB0_BASE_HI, mmCP_RB0_CNTL, mmCP_RB0_RPTR_ADDR, mmCP_RB0_RPTR_ADDR_HI, mmCP_RB0_WPTR, mmCP_RB0_WPTR_HI, mmCP_RB_DOORBELL_CONTROL, mmCP_RB_DOORBELL_RANGE_LOWER, mmCP_RB_DOORBELL_RANGE_UPPER, mmCP_RB_VMID, mmCP_RB_WPTR_POLL_ADDR_HI, mmCP_RB_WPTR_POLL_ADDR_LO, gem5::AMDGPUDevice::setDoorbellType(), setHqdActive(), setHqdIbCtrl(), setHqdPqBase(), setHqdPqBaseHi(), setHqdPqControl(), setHqdPqDoorbellCtrl(), setHqdPqPtr(), setHqdPqRptrReportAddr(), setHqdPqRptrReportAddrHi(), setHqdPqWptrHi(), setHqdPqWptrLo(), setHqdPqWptrPollAddr(), setHqdPqWptrPollAddrHi(), setHqdVmid(), setRbBaseHi(), setRbBaseLo(), setRbCntl(), setRbDoorbellCntrl(), setRbDoorbellRangeHi(), setRbDoorbellRangeLo(), setRbRptrAddrHi(), setRbRptrAddrLo(), setRbVmid(), setRbWptrHi(), setRbWptrLo(), setRbWptrPollAddrHi(), and setRbWptrPollAddrLo().
|
private |
Definition at line 67 of file pm4_packet_processor.hh.
Referenced by getIpId().
|
private |
Definition at line 68 of file pm4_packet_processor.hh.
Referenced by getMMIORange().
|
private |
Definition at line 54 of file pm4_packet_processor.hh.
Referenced by decodeHeader(), getGARTAddr(), mapProcess(), mapQueues(), newQueue(), processMQD(), processSDMAMQD(), releaseMemDone(), setGPUDevice(), setUconfigReg(), translate(), unmapAllQueues(), unmapQueues(), unserialize(), writeData(), and writeMMIO().
|
private |
Definition at line 59 of file pm4_packet_processor.hh.
Referenced by getKiqDoorbellOffset(), mapKiq(), PM4PacketProcessor(), setHqdActive(), setHqdIbCtrl(), setHqdPqBase(), setHqdPqBaseHi(), setHqdPqControl(), setHqdPqDoorbellCtrl(), setHqdPqPtr(), setHqdPqRptrReportAddr(), setHqdPqRptrReportAddrHi(), setHqdPqWptrPollAddr(), setHqdPqWptrPollAddrHi(), and setHqdVmid().
|
private |
Definition at line 60 of file pm4_packet_processor.hh.
Referenced by mapKiq().
|
private |
Definition at line 56 of file pm4_packet_processor.hh.
Referenced by getPqDoorbellOffset(), mapPq(), PM4PacketProcessor(), setRbBaseHi(), setRbBaseLo(), setRbCntl(), setRbDoorbellCntrl(), setRbDoorbellRangeHi(), setRbDoorbellRangeLo(), setRbRptrAddrHi(), setRbRptrAddrLo(), setRbVmid(), setRbWptrHi(), setRbWptrLo(), setRbWptrPollAddrHi(), and setRbWptrPollAddrLo().
|
private |
Definition at line 57 of file pm4_packet_processor.hh.
Referenced by mapPq().
|
private |
Definition at line 63 of file pm4_packet_processor.hh.
Referenced by newQueue(), serialize(), unmapAllQueues(), and unserialize().
|
private |
Definition at line 65 of file pm4_packet_processor.hh.
Referenced by getQueue(), newQueue(), processMQD(), and updateReadIndex().