sdfg::tiles::AxisSchedule Class Reference

Target-neutral cooperation facts derived from one loop's schedule. More...

#include <sdfg/tiles/tile.h>

Public Member Functions
	AxisSchedule ()=default
	AxisSchedule (Level level, Space space, bool has_scratchpad, unsigned spatial_axis=0, symbolic::Integer parallel_size=symbolic::integer(0), bool needs_sync=false, symbolic::Integer lanes=symbolic::integer(1))
bool	has_scratchpad () const
Level	level () const
Space	space () const
unsigned	spatial_axis () const
const symbolic::Integer &	parallel_size () const
bool	needs_sync () const
const symbolic::Integer &	lanes () const

Static Public Member Functions
static std::optional< AxisSchedule >	classify (const structured_control_flow::ScheduleType &sched)
	Classify a loop's schedule, or std::nullopt for a sequential loop (which does not shape storage).
static std::optional< Level >	classify_level (const structured_control_flow::ScheduleType &sched)
	The cooperation Level a target assigns to sched, or std::nullopt when no registered target treats it as a scratchpad (device-parallel) schedule (a sequential or host-thread loop).
static bool	drives_cooperative_copy (const structured_control_flow::ScheduleType &sched)
	True when sched is a group-level schedule whose parallel threads can drive a cooperative shared-memory staging copy (a genuine offload schedule).

Detailed Description

Target-neutral cooperation facts derived from one loop's schedule.

Constructor & Destructor Documentation

AxisSchedule() [1/2]

sdfg::tiles::AxisSchedule::AxisSchedule ( )

default

AxisSchedule() [2/2]

sdfg::tiles::AxisSchedule::AxisSchedule	(	Level	level,
		Space	space,
		bool	has_scratchpad,
		unsigned	spatial_axis = 0,
		symbolic::Integer	parallel_size = symbolic::integer(0),
		bool	needs_sync = false,
		symbolic::Integer	lanes = symbolic::integer(1) )

Member Function Documentation

classify()

std::optional< AxisSchedule > sdfg::tiles::AxisSchedule::classify ( const structured_control_flow::ScheduleType & sched )

static

Classify a loop's schedule, or std::nullopt for a sequential loop (which does not shape storage).

New offload schedule types extend this factory.

classify_level()

std::optional< Level > sdfg::tiles::AxisSchedule::classify_level ( const structured_control_flow::ScheduleType & sched )

static

The cooperation Level a target assigns to sched, or std::nullopt when no registered target treats it as a scratchpad (device-parallel) schedule (a sequential or host-thread loop).

Neutral schedule test: == Level::Group means group-cooperative, == Level::Device means device-wide.

drives_cooperative_copy()

bool sdfg::tiles::AxisSchedule::drives_cooperative_copy ( const structured_control_flow::ScheduleType & sched )

static

True when sched is a group-level schedule whose parallel threads can drive a cooperative shared-memory staging copy (a genuine offload schedule).

Distinguishes real offload block schedules from legacy fused whole-kernel schedules, which cooperate at group level but cannot host a separate copy Map.

has_scratchpad()

bool sdfg::tiles::AxisSchedule::has_scratchpad ( ) const

inline

lanes()

const symbolic::Integer & sdfg::tiles::AxisSchedule::lanes ( ) const

inline

level()

Level sdfg::tiles::AxisSchedule::level ( ) const

inline

needs_sync()

bool sdfg::tiles::AxisSchedule::needs_sync ( ) const

inline

parallel_size()

const symbolic::Integer & sdfg::tiles::AxisSchedule::parallel_size ( ) const

inline

space()

Space sdfg::tiles::AxisSchedule::space ( ) const

inline

spatial_axis()

unsigned sdfg::tiles::AxisSchedule::spatial_axis ( ) const

inline

---

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

• opt/include/sdfg/tiles/tile.h