tvm::tir::InstructionKindNode Class Reference

Kind of an instruction, e.g. Split, Reorder, etc. Besides the name, every kind of instruction has its own properties, including: 1) A boolean indicating if the instruction is pure, i.e. change nothing in the schedule state 2) A functor that applies the instruction to a TensorIR schedule 3) A functor that converts the instruction to a statement in python syntax 4) A functor that serialize its attributes to JSON 5) A functor that deserialize its attributes from JSON. More...

#include <instruction.h>

Inheritance diagram for tvm::tir::InstructionKindNode:

Collaboration diagram for tvm::tir::InstructionKindNode:

Public Member Functions
bool	IsPostproc () const
	Checks if the instruction kind is EnterPostproc. More...
	TVM_FFI_DECLARE_OBJECT_INFO_FINAL ("tir.InstructionKind", InstructionKindNode, runtime::Object)

Static Public Member Functions
static void	RegisterReflection ()

Public Attributes
ffi::String	name
	The name of a kind of instructions. More...
bool	is_pure {false}
	Indicates if the instruction is pure, i.e. removing it alone doesn't mutate the schedule state. For example, the instruction GetBlock is pure because it changes nothing, while ComputeInline is not because removing it leads to a different resulting schedule. More...
FInstructionApply	f_apply_to_schedule {nullptr}
	A functor that applies the instruction to a TensorIR schedule. More...
FInstructionAsPython	f_as_python {nullptr}
	A functor that converts the instruction to a statement in python syntax. More...
FInstructionAttrsAsJSON	f_attrs_as_json {nullptr}
	A functor that serialize its attributes to JSON. More...
FInstructionAttrsFromJSON	f_attrs_from_json {nullptr}
	A functor that deserialize its attributes from JSON. More...

Detailed Description

Kind of an instruction, e.g. Split, Reorder, etc. Besides the name, every kind of instruction has its own properties, including: 1) A boolean indicating if the instruction is pure, i.e. change nothing in the schedule state 2) A functor that applies the instruction to a TensorIR schedule 3) A functor that converts the instruction to a statement in python syntax 4) A functor that serialize its attributes to JSON 5) A functor that deserialize its attributes from JSON.

Unlike tvm::OpNode, InstructionKindNode doesn't support unstructured properties, mainly because there is no such usecase yet to add any other property.

Member Function Documentation

IsPostproc()

bool tvm::tir::InstructionKindNode::IsPostproc

(

)

const

Checks if the instruction kind is EnterPostproc.

RegisterReflection()

static void tvm::tir::InstructionKindNode::RegisterReflection ( )

inlinestatic

TVM_FFI_DECLARE_OBJECT_INFO_FINAL()

tvm::tir::InstructionKindNode::TVM_FFI_DECLARE_OBJECT_INFO_FINAL	(	"tir.InstructionKind"	,
		InstructionKindNode	,
		runtime::Object
	)

Member Data Documentation

f_apply_to_schedule

FInstructionApply tvm::tir::InstructionKindNode::f_apply_to_schedule {nullptr}

A functor that applies the instruction to a TensorIR schedule.

f_as_python

FInstructionAsPython tvm::tir::InstructionKindNode::f_as_python {nullptr}

A functor that converts the instruction to a statement in python syntax.

f_attrs_as_json

FInstructionAttrsAsJSON tvm::tir::InstructionKindNode::f_attrs_as_json {nullptr}

A functor that serialize its attributes to JSON.

Note

If the functor is null, it means no conversion is needed

f_attrs_from_json

FInstructionAttrsFromJSON tvm::tir::InstructionKindNode::f_attrs_from_json {nullptr}

A functor that deserialize its attributes from JSON.

Note

If the functor is null, it means no conversion is needed

is_pure

bool tvm::tir::InstructionKindNode::is_pure {false}

Indicates if the instruction is pure, i.e. removing it alone doesn't mutate the schedule state. For example, the instruction GetBlock is pure because it changes nothing, while ComputeInline is not because removing it leads to a different resulting schedule.

name

ffi::String tvm::tir::InstructionKindNode::name

The name of a kind of instructions.

---

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

• include/tvm/tir/schedule/instruction.h