HiCR::backend::hwloc::ComputeResource Class Reference

#include <computeResource.hpp>

Inheritance diagram for HiCR::backend::hwloc::ComputeResource:

Public Types
using	logicalProcessorId_t = unsigned int
using	physicalProcessorId_t = unsigned int
using	numaAffinity_t = unsigned int

Public Member Functions
	ComputeResource (hwloc_topology_t topology, const logicalProcessorId_t logicalProcessorId)
	ComputeResource (const logicalProcessorId_t logicalProcessorId, const physicalProcessorId_t physicalProcessorId, const numaAffinity_t numaAffinity, std::unordered_set< std::shared_ptr< backend::hwloc::Cache > > caches)
	~ComputeResource () override=default
	ComputeResource ()=default
__INLINE__ std::string	getType () const override
__INLINE__ logicalProcessorId_t	getProcessorId () const
__INLINE__ physicalProcessorId_t	getPhysicalProcessorId () const
	ComputeResource (const nlohmann::json &input)
Public Member Functions inherited from HiCR::ComputeResource
__INLINE__ nlohmann::json	serialize () const
__INLINE__ void	deserialize (const nlohmann::json &input)

Static Public Member Functions
static __INLINE__ void	detectThreadPUs (hwloc_topology_t topology, hwloc_obj_t obj, int depth, std::vector< logicalProcessorId_t > &threadPUs)
static __INLINE__ physicalProcessorId_t	detectPhysicalProcessorId (hwloc_topology_t topology, const logicalProcessorId_t logicalProcessorId)
static __INLINE__ numaAffinity_t	detectCoreNUMAffinity (hwloc_topology_t topology, const logicalProcessorId_t logicalProcessorId)
static __INLINE__ std::unordered_set< std::shared_ptr< backend::hwloc::Cache > >	detectCpuCaches (hwloc_topology_t topology, const logicalProcessorId_t logicalProcessorId)
static __INLINE__ numaAffinity_t	getCpuNumaAffinity (hwloc_topology_t topology, const logicalProcessorId_t logicalProcessorId)

Protected Member Functions
__INLINE__ void	serializeImpl (nlohmann::json &output) const override
__INLINE__ void	deserializeImpl (const nlohmann::json &input) override

Detailed Description

This class represents a compute resource, visible by HWLoc. That is, a CPU processing unit (core or hyperthread) with information about caches and locality.

Member Typedef Documentation

logicalProcessorId_t

using HiCR::backend::hwloc::ComputeResource::logicalProcessorId_t = unsigned int

System-given logical processor (core or hyperthread) identifier that this class instance represents

numaAffinity_t

using HiCR::backend::hwloc::ComputeResource::numaAffinity_t = unsigned int

System-given NUMA affinity identifier

physicalProcessorId_t

using HiCR::backend::hwloc::ComputeResource::physicalProcessorId_t = unsigned int

System-given physical processor identifier that this class instance represents

Constructor & Destructor Documentation

ComputeResource() [1/4]

HiCR::backend::hwloc::ComputeResource::ComputeResource ( hwloc_topology_t  topology, const logicalProcessorId_t  logicalProcessorId  )

inline

Constructor for the compute resource class of the hwloc backend

Parameters

topology	HWLoc topology object for discovery
logicalProcessorId	Os-determied core affinity assigned to this compute resource

ComputeResource() [2/4]

HiCR::backend::hwloc::ComputeResource::ComputeResource ( const logicalProcessorId_t  logicalProcessorId, const physicalProcessorId_t  physicalProcessorId, const numaAffinity_t  numaAffinity, std::unordered_set< std::shared_ptr< backend::hwloc::Cache > >  caches  )

inline

Constructor for the compute resource class of the hwloc backend

Parameters

[in]	logicalProcessorId	Unique identifier for the core assigned to this compute resource
[in]	numaAffinity	The NUMA domain associated to this core
[in]	caches	The set of caches contained to or accessible by this core
[in]	physicalProcessorId	The identifier of the physical core as assigned by the OS

~ComputeResource()

HiCR::backend::hwloc::ComputeResource::~ComputeResource ( )

overridevirtualdefault

Default destructor

Reimplemented from HiCR::ComputeResource.

ComputeResource() [3/4]

HiCR::backend::hwloc::ComputeResource::ComputeResource ( )

default

Default constructor for serialization/deserialization purposes

ComputeResource() [4/4]

HiCR::backend::hwloc::ComputeResource::ComputeResource ( const nlohmann::json &  input )

inline

Deserializing constructor

The instance created will contain all information, if successful in deserializing it, corresponding to the passed processing unit This instance should NOT be used for anything else than reporting/printing the contained resources

Parameters

[in]

input

A JSON-encoded serialized processing unit information

Member Function Documentation

deserializeImpl()

__INLINE__ void HiCR::backend::hwloc::ComputeResource::deserializeImpl ( const nlohmann::json &  input )

inlineoverrideprotectedvirtual

Backend-specific implementation of the deserialize function

Parameters

[in]

input

Serialized compute resource information corresponding to the specific backend's topology manager

Implements HiCR::ComputeResource.

detectCoreNUMAffinity()

static __INLINE__ numaAffinity_t HiCR::backend::hwloc::ComputeResource::detectCoreNUMAffinity ( hwloc_topology_t  topology, const logicalProcessorId_t  logicalProcessorId  )

inlinestatic

Uses HWloc to discover the NUMA node associated with a given logical processor ID

Parameters

[in]	topology	An HWLoc topology object, already initialized
[in]	logicalProcessorId	The ID of the processor we are doing the search for

Returns

The ID of the associated memory space

detectCpuCaches()

static __INLINE__ std::unordered_set< std::shared_ptr< backend::hwloc::Cache > > HiCR::backend::hwloc::ComputeResource::detectCpuCaches ( hwloc_topology_t  topology, const logicalProcessorId_t  logicalProcessorId  )

inlinestatic

Uses HWloc to discover all caches associated with a given logical processor ID

Parameters

[in]	topology	An HWLoc topology object, already initialized
[in]	logicalProcessorId	The logical ID of the processor we are doing the search for

Returns

A vector of (type,size) entries, where type is a string describing the cache and size is the respective size in Bytes The 'type' string is has the following form: "Level <I/D/U> <P/S> <associated IDs>", where: Level: may be "L1", "L2, "L3" I/D/U: may be "Instruction", "Data", "Unified" P/S: may be "Private" or "Shared" associated IDs: (optional, for Shared cache) a list of core IDs, e.g. "0 1 2 3"

detectPhysicalProcessorId()

static __INLINE__ physicalProcessorId_t HiCR::backend::hwloc::ComputeResource::detectPhysicalProcessorId ( hwloc_topology_t  topology, const logicalProcessorId_t  logicalProcessorId  )

inlinestatic

Uses HWloc to discover the (physical) processor ID, associated with a given logical processor ID

Parameters

[in]	topology	An HWLoc topology object, already initialized
[in]	logicalProcessorId	The logical ID of the processor we are doing the search for

Returns

The ID of the associated physical identifier related to the passed logical processor id

detectThreadPUs()

static __INLINE__ void HiCR::backend::hwloc::ComputeResource::detectThreadPUs ( hwloc_topology_t  topology, hwloc_obj_t  obj, int  depth, std::vector< logicalProcessorId_t > &  threadPUs  )

inlinestatic

Uses HWloc to recursively (tree-like) identify the host's basic processing units (PUs)

Parameters

[in]	topology	An HWLoc topology object, already initialized
[in]	obj	The root HWLoc object for the start of the exploration tree at every recursion level
[in]	depth	Stores the current exploration depth level, necessary to return only the processing units at the leaf level
[out]	threadPUs	Storage for the found procesing units

getCpuNumaAffinity()

static __INLINE__ numaAffinity_t HiCR::backend::hwloc::ComputeResource::getCpuNumaAffinity ( hwloc_topology_t  topology, const logicalProcessorId_t  logicalProcessorId  )

inlinestatic

Uses HWloc to discover the NUMA node associated with a given logical processor ID

Parameters

[in]	topology	HWLoc topology object
[in]	logicalProcessorId	The ID of the processor we are doing the search for

Returns

The ID of the associated memory space

getPhysicalProcessorId()

__INLINE__ physicalProcessorId_t HiCR::backend::hwloc::ComputeResource::getPhysicalProcessorId ( ) const

inline

Obtains the Core ID of the CPU; in non SMT systems that will be the actual id; in SMT it is the id of the actual core the thread belongs to.

Returns

The physical ID of the hardware Core

getProcessorId()

__INLINE__ logicalProcessorId_t HiCR::backend::hwloc::ComputeResource::getProcessorId ( ) const

inline

Function to return the compute resource processor id

Returns

The processor id

getType()

__INLINE__ std::string HiCR::backend::hwloc::ComputeResource::getType ( ) const

inlineoverridevirtual

Indicates what type of compute unit is contained in this instance

Returns

A string containing a human-readable description of the compute resource type

Implements HiCR::ComputeResource.

serializeImpl()

__INLINE__ void HiCR::backend::hwloc::ComputeResource::serializeImpl ( nlohmann::json &  output ) const

inlineoverrideprotectedvirtual

Backend-specific implemtation of the serialize function that allows adding more information than the one provided by default by HiCR

Parameters

[out]

output

JSON-formatted serialized compute resource information

Implements HiCR::ComputeResource.

---

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

• /home/runner/work/HiCR/HiCR/include/hicr/backends/hwloc/computeResource.hpp