Represents a physical accelerated computing device. More...

#include <hc.hpp>

Collaboration diagram for hc::accelerator:

Public Member Functions
	accelerator ()
	Constructs a new accelerator object that represents the default accelerator. More...

	accelerator (const std::wstring &path)
	Constructs a new accelerator object that represents the physical device named by the "path" argument. More...

	accelerator (const accelerator &other)
	Copy constructs an accelerator object. More...

accelerator &	operator= (const accelerator &other)
	Assigns an accelerator object to "this" accelerator object and returns a reference to "this" object. More...

accelerator_view	get_default_view () const
	Returns the default accelerator_view associated with the accelerator. More...

accelerator_view	create_view (execute_order order=execute_in_order, queuing_mode mode=queuing_mode_automatic)
	Creates and returns a new accelerator view on the accelerator with the supplied queuing mode. More...

bool	operator== (const accelerator &other) const
	Compares "this" accelerator with the passed accelerator object to determine if they represent the same underlying device. More...

bool	operator!= (const accelerator &other) const
	Compares "this" accelerator with the passed accelerator object to determine if they represent different devices. More...

bool	set_default_cpu_access_type (access_type type)
	Sets the default_cpu_access_type for this accelerator. More...

std::wstring	get_device_path () const
	Returns a system-wide unique device instance path that matches the "Device Instance Path" property for the device in Device Manager, or one of the predefined path constants cpu_accelerator.

std::wstring	get_description () const
	Returns a short textual description of the accelerator device.

unsigned int	get_version () const
	Returns a 32-bit unsigned integer representing the version number of this accelerator. More...

bool	get_has_display () const
	This property indicates that the accelerator may be shared by (and thus have interference from) the operating system or other system software components for rendering purposes. More...

size_t	get_dedicated_memory () const
	Returns the amount of dedicated memory (in KB) on an accelerator device. More...

bool	get_supports_double_precision () const
	Returns a Boolean value indicating whether this accelerator supports double-precision (double) computations. More...

bool	get_supports_limited_double_precision () const
	Returns a boolean value indicating whether the accelerator has limited double precision support (excludes double division, precise_math functions, int to double, double to int conversions) for a parallel_for_each kernel.

bool	get_is_debug () const
	Returns a boolean value indicating whether the accelerator supports debugging.

bool	get_is_emulated () const
	Returns a boolean value indicating whether the accelerator is emulated. More...

bool	get_supports_cpu_shared_memory () const
	Returns a boolean value indicating whether the accelerator supports memory accessible both by the accelerator and the CPU.

access_type	get_default_cpu_access_type () const
	Get the default cpu access_type for buffers created on this accelerator.

size_t	get_max_tile_static_size ()
	Returns the maximum size of tile static area available on this accelerator.

std::vector< accelerator_view >	get_all_views ()
	Returns a vector of all accelerator_view associated with this accelerator.

void *	get_hsa_am_region () const
	Returns an opaque handle which points to the AM region on the HSA agent. More...

void *	get_hsa_am_system_region () const
	Returns an opaque handle which points to the AM system region on the HSA agent. More...

void *	get_hsa_am_finegrained_system_region () const
	Returns an opaque handle which points to the AM system region on the HSA agent. More...

void *	get_hsa_kernarg_region () const
	Returns an opaque handle which points to the Kernarg region on the HSA agent. More...

bool	is_hsa_accelerator () const
	Returns if the accelerator is based on HSA.

hcAgentProfile	get_profile () const
	Returns the profile the accelerator. More...

void	memcpy_symbol (const char symbolName, void hostptr, size_t count, size_t offset=0, hcCommandKind kind=hcMemcpyHostToDevice)

void	memcpy_symbol (void symbolAddr, void hostptr, size_t count, size_t offset=0, hcCommandKind kind=hcMemcpyHostToDevice)

void *	get_symbol_address (const char *symbolName)

void *	get_hsa_agent () const
	Returns an opaque handle which points to the underlying HSA agent. More...

bool	get_is_peer (const accelerator &other) const
	Check if `other` is peer of this accelerator. More...

std::vector< accelerator >	get_peers () const
	Return a std::vector of this accelerator's peers. More...

unsigned int	get_cu_count () const
	Return the compute unit count of the accelerator.

int	get_seqnum () const
	Return the unique integer sequence-number for the accelerator. More...

bool	has_cpu_accessible_am ()
	Return true if the accelerator's memory can be mapped into the CPU's address space, and the CPU is allowed to access the memory directly with CPU memory operations. More...

Kalmar::KalmarDevice *	get_dev_ptr () const

Static Public Member Functions
static std::vector< accelerator >	get_all ()
	Returns a std::vector of accelerator objects (in no specific order) representing all accelerators that are available, including reference accelerators and WARP accelerators if available. More...

static bool	set_default (const std::wstring &path)
	Sets the default accelerator to the device path identified by the "path" argument. More...

static accelerator_view	get_auto_selection_view ()
	Returns an accelerator_view which when passed as the first argument to a parallel_for_each call causes the runtime to automatically select the target accelerator_view for executing the parallel_for_each kernel. More...

Friends
class	accelerator_view

Detailed Description

Represents a physical accelerated computing device.

An object of this type can be created by enumerating the available devices, or getting the default device.

Constructor & Destructor Documentation

hc::accelerator::accelerator ( )

inline

Constructs a new accelerator object that represents the default accelerator.

This is equivalent to calling the constructor

accelerator(accelerator::default_accelerator)

The actual accelerator chosen as the default can be affected by calling accelerator::set_default().

hc::accelerator::accelerator ( const std::wstring & path )

inlineexplicit

Constructs a new accelerator object that represents the physical device named by the "path" argument.

If the path represents an unknown or unsupported device, an exception will be thrown.

The path can be one of the following:

accelerator::default_accelerator (or L"default"), which represents the path of the fastest accelerator available, as chosen by the runtime.
accelerator::cpu_accelerator (or L"cpu"), which represents the CPU. Note that parallel_for_each shall not be invoked over this accelerator.
A valid device path that uniquely identifies a hardware accelerator available on the host system.

Parameters

[in] path The device path of this accelerator.

hc::accelerator::accelerator ( const accelerator & other )

inline

Copy constructs an accelerator object.

This function does a shallow copy with the newly created accelerator object pointing to the same underlying device as the passed accelerator parameter.

Parameters

[in] other The accelerator object to be copied.

Member Function Documentation

accelerator_view hc::accelerator::create_view	(	execute_order	order = `execute_in_order`,
		queuing_mode	mode = `queuing_mode_automatic`
	)

inline

Creates and returns a new accelerator view on the accelerator with the supplied queuing mode.

Parameters

[in] qmode The queuing mode of the accelerator_view to be created. See "Queuing Mode". The default value would be queueing_mdoe_automatic if not specified.

static std::vector<accelerator> hc::accelerator::get_all ( )

inlinestatic

Returns a std::vector of accelerator objects (in no specific order) representing all accelerators that are available, including reference accelerators and WARP accelerators if available.

Returns: A vector of accelerators.

static accelerator_view hc::accelerator::get_auto_selection_view ( )

inlinestatic

Returns an accelerator_view which when passed as the first argument to a parallel_for_each call causes the runtime to automatically select the target accelerator_view for executing the parallel_for_each kernel.

In other words, a parallel_for_each invocation with the accelerator_view returned by get_auto_selection_view() is the same as a parallel_for_each invocation without an accelerator_view argument.

For all other purposes, the accelerator_view returned by get_auto_selection_view() behaves the same as the default accelerator_view of the default accelerator (aka accelerator().get_default_view() ).

Returns: An accelerator_view than can be used to indicate auto selection of the target for a parallel_for_each execution.

size_t hc::accelerator::get_dedicated_memory ( ) const

inline

Returns the amount of dedicated memory (in KB) on an accelerator device.

There is no guarantee that this amount of memory is actually available to use.

accelerator_view hc::accelerator::get_default_view ( ) const

inline

Returns the default accelerator_view associated with the accelerator.

The queuing_mode of the default accelerator_view is queuing_mode_automatic.

Returns: The default accelerator_view object associated with the accelerator.

bool hc::accelerator::get_has_display ( ) const

inline

This property indicates that the accelerator may be shared by (and thus have interference from) the operating system or other system software components for rendering purposes.

A C++ AMP implementation may set this property to false should such interference not be applicable for a particular accelerator.

void* hc::accelerator::get_hsa_agent ( ) const

inline

Returns an opaque handle which points to the underlying HSA agent.

Returns: An opaque handle of the underlying HSA agent, if the accelerator is based on HSA. NULL otherwise.

void* hc::accelerator::get_hsa_am_finegrained_system_region ( ) const

inline

Returns an opaque handle which points to the AM system region on the HSA agent.

This region can be used to allocate finegrained system memory which is accessible from the specified accelerator.

Returns: An opaque handle of the region, if the accelerator is based on HSA. NULL otherwise.

void* hc::accelerator::get_hsa_am_region ( ) const

inline

Returns an opaque handle which points to the AM region on the HSA agent.

This region can be used to allocate accelerator memory which is accessible from the specified accelerator.

Returns: An opaque handle of the region, if the accelerator is based on HSA. NULL otherwise.

void* hc::accelerator::get_hsa_am_system_region ( ) const

inline

Returns an opaque handle which points to the AM system region on the HSA agent.

This region can be used to allocate system memory which is accessible from the specified accelerator.

Returns: An opaque handle of the region, if the accelerator is based on HSA. NULL otherwise.

void* hc::accelerator::get_hsa_kernarg_region ( ) const

inline

Returns an opaque handle which points to the Kernarg region on the HSA agent.

Returns: An opaque handle of the region, if the accelerator is based on HSA. NULL otherwise.

bool hc::accelerator::get_is_emulated ( ) const

inline

Returns a boolean value indicating whether the accelerator is emulated.

This is true, for example, with the reference, WARP, and CPU accelerators.

bool hc::accelerator::get_is_peer ( const accelerator & other ) const

inline

Check if other is peer of this accelerator.

Returns: true if other can access this accelerator's device memory pool or false if not. The acceleratos is not its own peer.

std::vector<accelerator> hc::accelerator::get_peers ( ) const

inline

Return a std::vector of this accelerator's peers.

peer is other accelerator which can access this accelerator's device memory using map_to_peer family of APIs.

hcAgentProfile hc::accelerator::get_profile ( ) const

inline

Returns the profile the accelerator.

hcAgentProfileNone in case the accelerator is not based on HSA.
hcAgentProfileBase in case the accelerator is of HSA Base Profile.
hcAgentProfileFull in case the accelerator is of HSA Full Profile.

int hc::accelerator::get_seqnum ( ) const

inline

Return the unique integer sequence-number for the accelerator.

Sequence-numbers are assigned in monotonically increasing order starting with 0.

bool hc::accelerator::get_supports_double_precision ( ) const

inline

Returns a Boolean value indicating whether this accelerator supports double-precision (double) computations.

When this returns true, supports_limited_double_precision also returns true.

unsigned int hc::accelerator::get_version ( ) const

inline

Returns a 32-bit unsigned integer representing the version number of this accelerator.

The format of the integer is major.minor, where the major version number is in the high-order 16 bits, and the minor version number is in the low-order bits.

bool hc::accelerator::has_cpu_accessible_am ( )

inline

Return true if the accelerator's memory can be mapped into the CPU's address space, and the CPU is allowed to access the memory directly with CPU memory operations.

Typically this is enabled with "large BAR" or "resizeable BAR" address mapping.

bool hc::accelerator::operator!= ( const accelerator & other ) const

inline

Compares "this" accelerator with the passed accelerator object to determine if they represent different devices.

Parameters

[in] other The accelerator object to be compared against.

Returns: A boolean value indicating whether the passed accelerator object is different from "this" accelerator.

accelerator& hc::accelerator::operator= ( const accelerator & other )

inline

Assigns an accelerator object to "this" accelerator object and returns a reference to "this" object.

This function does a shallow assignment with the newly created accelerator object pointing to the same underlying device as the passed accelerator parameter.

Parameters

other The accelerator object to be assigned from.

Returns: A reference to "this" accelerator object.

bool hc::accelerator::operator== ( const accelerator & other ) const

inline

Compares "this" accelerator with the passed accelerator object to determine if they represent the same underlying device.

Parameters

[in] other The accelerator object to be compared against.

Returns: A boolean value indicating whether the passed accelerator object is same as "this" accelerator.

static bool hc::accelerator::set_default ( const std::wstring & path )

inlinestatic

Sets the default accelerator to the device path identified by the "path" argument.

See the constructor accelerator(const std::wstring& path) for a description of the allowable path strings.

This establishes a process-wide default accelerator and influences all subsequent operations that might use a default accelerator.

Parameters

[in] path The device path of the default accelerator.

Returns: A Boolean flag indicating whether the default was set. If the default has already been set for this process, this value will be false, and the function will have no effect.

bool hc::accelerator::set_default_cpu_access_type ( access_type type )

inline

Sets the default_cpu_access_type for this accelerator.

The default_cpu_access_type is used for arrays created on this accelerator or for implicit array_view memory allocations accessed on this this accelerator.

This method only succeeds if the default_cpu_access_type for the accelerator has not already been overriden by a previous call to this method and the runtime selected default_cpu_access_type for this accelerator has not yet been used for allocating an array or for an implicit array_view memory allocation on this accelerator.

Parameters

[in] default_cpu_access_type The default cpu access_type to be used for array/array_view memory allocations on this accelerator.

Returns: A boolean value indicating if the default cpu access_type for the accelerator was successfully set.

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

/home/scchan/code/hcc/include/hc.hpp

Public Member Functions

Static Public Member Functions

Friends

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation