#include <gpu_compute_driver.hh>

Inheritance diagram for GPUComputeDriver:

Public Types
typedef GPUComputeDriverParams	Params

Public Types inherited from SimObject
typedef SimObjectParams	Params

Public Member Functions
	GPUComputeDriver (Params *p)

int	ioctl (ThreadContext *tc, unsigned req, Addr ioc_buf) override
	Abstract method, invoked when the user program calls ioctl() on the file descriptor returned by a previous open(). More...

Public Member Functions inherited from HSADriver
	HSADriver (HSADriverParams *p)

int	open (ThreadContext *tc, int mode, int flags)
	Create an FD entry for the KFD inside of the owning process. More...

Addr	mmap (ThreadContext *tc, Addr start, uint64_t length, int prot, int tgtFlags, int tgtFd, int offset)
	Currently, mmap() will simply setup a mapping for the associated device's packet processor's doorbells. More...

Public Member Functions inherited from EmulatedDriver
	EmulatedDriver (EmulatedDriverParams *p)

bool	match (const std::string &s) const
	Check for a match with this driver's filename. More...

Public Member Functions inherited from SimObject
const Params *	params () const

	SimObject (const Params *_params)

virtual	~SimObject ()

virtual const std::string	name () const

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...

ProbeManager *	getProbeManager ()
	Get the probe manager for this object. More...

virtual Port &	getPort (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 EventManager
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. 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 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 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 Stats::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 (Stats::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 Info *	resolveStat (std::string name) const
	Resolve a stat by its name within this group. More...

	Group ()=delete

	Group (const Group &)=delete

Group &	operator= (const Group &)=delete

Private Member Functions
Addr	gpuVmApeBase (int gpuNum) const
	The aperture (APE) base/limit pairs are set statically at startup by the real KFD. More...

Addr	gpuVmApeLimit (Addr apeBase) const

Addr	scratchApeBase (int gpuNum) const

Addr	scratchApeLimit (Addr apeBase) const

Addr	ldsApeBase (int gpuNum) const

Addr	ldsApeLimit (Addr apeBase) const

Additional Inherited Members
Static Public Member Functions inherited from SimObject
static void	serializeAll (CheckpointOut &cp)
	Serialize all SimObjects in the system. More...

static SimObject *	find (const char *name)
	Find the SimObject with the given name and return a pointer to it. More...

Static Public Member Functions inherited from Serializable
static const std::string &	currentSection ()
	Gets the fully-qualified name of the active section. More...

static void	serializeAll (const std::string &cpt_dir)
	Serializes all the SimObjects. More...

static void	unserializeGlobals (CheckpointIn &cp)

Protected Member Functions inherited from HSADriver
void	allocateQueue (ThreadContext *tc, Addr ioc_buf)
	Forward relevant parameters to packet processor; queueID is used to link doorbell. More...

Protected Member Functions inherited from Drainable
	Drainable ()

virtual	~Drainable ()

virtual void	drainResume ()
	Resume execution after a successful drain. More...

void	signalDrainDone () const
	Signal that an object is drained. More...

Protected Attributes inherited from HSADriver
HSADevice *	device
	HSA agent (device) that is controled by this driver. More...

uint32_t	queueId

Protected Attributes inherited from EmulatedDriver
const std::string &	filename
	filename for opening this driver (under /dev) More...

Protected Attributes inherited from SimObject
const SimObjectParams *	_params
	Cached copy of the object parameters. More...

Protected Attributes inherited from EventManager
EventQueue *	eventq
	A pointer to this object's event queue. More...

Detailed Description

Definition at line 52 of file gpu_compute_driver.hh.

Member Typedef Documentation

◆ Params

typedef GPUComputeDriverParams GPUComputeDriver::Params

Definition at line 55 of file gpu_compute_driver.hh.

Constructor & Destructor Documentation

◆ GPUComputeDriver()

GPUComputeDriver::GPUComputeDriver ( Params * p )

Definition at line 47 of file gpu_compute_driver.cc.

References DPRINTF.

Member Function Documentation

◆ gpuVmApeBase()

Addr GPUComputeDriver::gpuVmApeBase ( int gpuNum ) const

private

The aperture (APE) base/limit pairs are set statically at startup by the real KFD.

AMD x86_64 CPUs only use the areas in the 64b address space where VA[63:47] == 0x1ffff or VA[63:47] = 0. These methods generate the APE base/limit pairs in exactly the same way as the real KFD does, which ensures these APEs do not fall into the CPU's address space

see the macros in the KFD driver in the ROCm Linux kernel source:

drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c

Definition at line 379 of file gpu_compute_driver.cc.

References X86ISA::L.

Referenced by ioctl().

◆ gpuVmApeLimit()

Addr GPUComputeDriver::gpuVmApeLimit ( Addr apeBase ) const

private

Definition at line 385 of file gpu_compute_driver.cc.

References X86ISA::L.

Referenced by ioctl().

◆ ioctl()

int GPUComputeDriver::ioctl	(	ThreadContext *	tc,
		unsigned	req,
		Addr	buf
	)

overridevirtual

Abstract method, invoked when the user program calls ioctl() on the file descriptor returned by a previous open().

The parameters are the same as those passed in to ioctlFunc() (q.v.).

Returns: The return code for the ioctl, or the negation of the errno (see the SyscallReturn class).

Derive all clock counters based on the tick. All device clocks are identical and perfectly in sync.

Set the GPUVM/LDS/Scratch APEs exactly as they are in the real driver, see the KFD driver in the ROCm Linux kernel source: drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c

While the GPU node numbers start at 0, we add 1 to force the count to start at 1. This is to ensure that the base/limit addresses are calculated correctly.

The CPU's 64b address space can only use the areas with VA[63:47] == 0x1ffff or VA[63:47] == 0, therefore we must ensure that the apertures do not fall in the CPU's address space.

Implements EmulatedDriver.

Definition at line 54 of file gpu_compute_driver.cc.

◆ ldsApeBase()

Addr GPUComputeDriver::ldsApeBase ( int gpuNum ) const

private

Definition at line 403 of file gpu_compute_driver.cc.

Referenced by ioctl().

◆ ldsApeLimit()

Addr GPUComputeDriver::ldsApeLimit ( Addr apeBase ) const

private

Definition at line 409 of file gpu_compute_driver.cc.

Referenced by ioctl().

◆ scratchApeBase()

Addr GPUComputeDriver::scratchApeBase ( int gpuNum ) const

private

Definition at line 391 of file gpu_compute_driver.cc.

References X86ISA::L.

Referenced by ioctl().

◆ scratchApeLimit()

Addr GPUComputeDriver::scratchApeLimit ( Addr apeBase ) const

private

Definition at line 397 of file gpu_compute_driver.cc.

Referenced by ioctl().

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

gpu-compute/gpu_compute_driver.hh
gpu-compute/gpu_compute_driver.cc

Public Types

Public Member Functions

Private Member Functions

Additional Inherited Members

Detailed Description

Member Typedef Documentation

◆ Params

Constructor & Destructor Documentation

◆ GPUComputeDriver()

Member Function Documentation

◆ gpuVmApeBase()

◆ gpuVmApeLimit()

◆ ioctl()

◆ ldsApeBase()

◆ ldsApeLimit()

◆ scratchApeBase()

◆ scratchApeLimit()