threadschedule::ThreadAffinity Class Reference

Manages a set of CPU indices to which a thread may be bound. More...

#include <scheduler_policy.hpp>

Public Member Functions
	ThreadAffinity ()
	ThreadAffinity (std::vector< int > const &cpus)
void	add_cpu (int cpu)
void	remove_cpu (int cpu)
auto	is_set (int cpu) const -> bool
auto	has_cpu (int cpu) const -> bool
void	clear ()
auto	get_cpus () const -> std::vector< int >
auto	native_handle () const -> cpu_set_t const &
auto	to_string () const -> std::string

Detailed Description

Manages a set of CPU indices to which a thread may be bound.

ThreadAffinity is a value-semantic type that represents a CPU affinity mask. It abstracts away the platform-specific details of cpu_set_t (Linux) and processor-group bitmasks (Windows).

Linux

Backed by a cpu_set_t. Supports CPU indices in the range [0, CPU_SETSIZE) (typically 0-1023). The native_handle() accessor provides a const cpu_set_t& for direct use with pthread_setaffinity_np or sched_setaffinity.

Windows

Backed by a 64-bit bitmask plus a processor group index (WORD). Windows organises logical processors into groups of up to 64. This class supports a single group at a time: the group is determined by the first CPU added via add_cpu(). Subsequent calls to add_cpu() for CPUs that belong to a different group are silently ignored. Use get_group() and get_mask() to retrieve the platform-native values for SetThreadGroupAffinity.

Thread safety

None. ThreadAffinity is a plain value type with no internal synchronisation. Concurrent reads are safe; concurrent mutation (or a read concurrent with a write) requires external locking.

Copyability / movability

Implicitly copyable and movable (compiler-generated special members).

Warning

On Windows, CPUs from different processor groups cannot be combined in a single ThreadAffinity instance. If you need cross-group affinity you must apply separate ThreadAffinity objects per group.

Definition at line 213 of file scheduler_policy.hpp.

Constructor & Destructor Documentation

ThreadAffinity() [1/2]

threadschedule::ThreadAffinity::ThreadAffinity ( )

inline

Definition at line 216 of file scheduler_policy.hpp.

Referenced by ThreadAffinity().

ThreadAffinity() [2/2]

threadschedule::ThreadAffinity::ThreadAffinity ( std::vector< int > const & cpus )

inlineexplicit

Definition at line 226 of file scheduler_policy.hpp.

References add_cpu(), and ThreadAffinity().

Member Function Documentation

add_cpu()

void threadschedule::ThreadAffinity::add_cpu ( int cpu )

inline

Definition at line 235 of file scheduler_policy.hpp.

Referenced by threadschedule::affinity_for_node(), and ThreadAffinity().

clear()

void threadschedule::ThreadAffinity::clear ( )

inline

Definition at line 297 of file scheduler_policy.hpp.

get_cpus()

auto threadschedule::ThreadAffinity::get_cpus ( ) const -> std::vector<int>

inlinenodiscard

Definition at line 306 of file scheduler_policy.hpp.

Referenced by to_string().

has_cpu()

auto threadschedule::ThreadAffinity::has_cpu ( int cpu ) const -> bool

inlinenodiscard

Definition at line 292 of file scheduler_policy.hpp.

References is_set().

is_set()

auto threadschedule::ThreadAffinity::is_set ( int cpu ) const -> bool

inlinenodiscard

Definition at line 279 of file scheduler_policy.hpp.

Referenced by has_cpu().

native_handle()

auto threadschedule::ThreadAffinity::native_handle ( ) const -> cpu_set_t const&

inlinenodiscard

Definition at line 343 of file scheduler_policy.hpp.

remove_cpu()

void threadschedule::ThreadAffinity::remove_cpu ( int cpu )

inline

Definition at line 260 of file scheduler_policy.hpp.

to_string()

auto threadschedule::ThreadAffinity::to_string ( ) const -> std::string

inlinenodiscard

Definition at line 349 of file scheduler_policy.hpp.

References get_cpus().

---

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

• scheduler_policy.hpp