mlir.dialects.transform.tune¶

Classes¶

`Type`
`Value`	Abstract base class for generic types.
`Operation`
`OpView`
`Attribute`
`ArrayAttr`
`StringAttr`
`F64Type`
`IntegerType`
`IntegerAttr`
`FloatAttr`
`BoolAttr`
`AlternativesOp`	This op represents a choice over which of its regions is to be used.
`AlternativesOpAdaptor`
`KnobOp`	Provides a representation for "tunables" within schedules.
`KnobOpAdaptor`
`KnobOp`	Provides a representation for "tunables" within schedules.
`AlternativesOp`	This op represents a choice over which of its regions is to be used.

Functions¶

`tune_alternatives`(→ Union[_ods_ir, _ods_ir, ...)
`tune_knob`(→ _ods_ir)
`knob`(result, name, options, *[, selected, loc, ip])
`alternatives`(results, name, num_alternatives, *[, ...])

Module Contents¶

class mlir.dialects.transform.tune.Type(cast_from_type: Type)¶

property _CAPIPtr: object¶: Gets a capsule wrapping the MlirType.

static _CAPICreate(arg: object, /) → Type¶: Creates a Type from a capsule wrapping MlirType.

static parse(asm: str, context: Context | None = None) → Type¶

Parses the assembly form of a type.

Returns a Type object or raises an MLIRError if the type cannot be parsed.

property context: Context¶: Context that owns the Type.

__eq__(arg: Type, /) → bool¶
__eq__(other: object | None) → bool: Compares type with non-type object (always returns False).

__hash__() → int¶: Returns the hash value of the Type.

dump() → None¶: Dumps a debug representation of the object to stderr.

__str__() → str¶: Returns the assembly form of the Type.

__repr__() → str¶: Returns a string representation of the Type.

maybe_downcast() → Type¶: Downcasts the Type to a more specific Type if possible.

property typeid: TypeID¶: Returns the TypeID of the Type, or raises ValueError if Type has no TypeID.

class mlir.dialects.transform.tune.Value(value: Value)¶

Bases: Generic[_T]

Abstract base class for generic types.

A generic type is typically declared by inheriting from this class parameterized with one or more type variables. For example, a generic mapping type might be defined as:

class Mapping(Generic[KT, VT]): def **getitem**(self, key: KT) -> VT: … # Etc.

This class can then be used as follows:

def lookup_name(mapping: Mapping[KT, VT], key: KT, default: VT) -> VT: try: return mapping[key] except KeyError: return default

property _CAPIPtr: object¶: Gets a capsule wrapping the MlirValue.

static _CAPICreate(arg: object, /) → Value¶: Creates a Value from a capsule wrapping MlirValue.

property context: Context¶: Context in which the value lives.

dump() → None¶: Dumps a debug representation of the object to stderr.

property owner: OpView | Block¶: Returns the owner of the value (Operation for results, Block for arguments).

property uses: OpOperandIterator¶: Returns an iterator over uses of this value.

__eq__(arg: Value, /) → bool¶
__eq__(arg: object, /) → bool: Compares value with non-value object (always returns False).

__hash__() → int¶: Returns the hash value of the value.

__str__() → str¶

Returns the string form of the value.

If the value is a block argument, this is the assembly form of its type and the position in the argument list. If the value is an operation result, this is equivalent to printing the operation that produced it.

get_name(use_local_scope: bool = False, use_name_loc_as_prefix: bool = False) → str¶
get_name(state: AsmState) → str: Returns the string form of value as an operand (i.e., the ValueID).

property type: _T¶: Returns the type of the value.

set_type(type: _T)¶: Sets the type of the value.

replace_all_uses_with(arg: Value, /) → None¶: Replace all uses of value with the new value, updating anything in the IR that uses self to use the other value instead.

replace_all_uses_except(with_: Value, exceptions: Operation) → None¶
replace_all_uses_except(with_: Value, exceptions: collections.abc.Sequence[Operation]) → None

maybe_downcast() → BlockArgument | OpResult | Value¶: Downcasts the Value to a more specific kind if possible.

property location: Location¶: Returns the source location of the value.

class mlir.dialects.transform.tune.Operation¶

Bases: _OperationBase

static create(name: str, results: collections.abc.Sequence[Type] | None = None, operands: collections.abc.Sequence[Value] | None = None, attributes: dict[str, Attribute] | None = None, successors: collections.abc.Sequence[Block] | None = None, regions: int = 0, loc: Location | None = None, ip: object | None = None, infer_type: bool = False) → Operation¶

Creates a new operation.

Parameters:

name – Operation name (e.g. dialect.operation).
results – Optional sequence of Type representing op result types.
operands – Optional operands of the operation.
attributes – Optional Dict of {str: Attribute}.
successors – Optional List of Block for the operation’s successors.
regions – Number of regions to create (default = 0).
location – Optional Location object (defaults to resolve from context manager).
ip – Optional InsertionPoint (defaults to resolve from context manager or set to False to disable insertion, even with an insertion point set in the context manager).
infer_type – Whether to infer result types (default = False).

Returns:

A new detached Operation object. Detached operations can be added to blocks, which causes them to become attached.

static parse(source: str, *, source_name: str = '', context: Context | None = None) → OpView¶: Parses an operation. Supports both text assembly format and binary bytecode format.

property _CAPIPtr: object¶: Gets a capsule wrapping the MlirOperation.

static _CAPICreate(arg: object, /) → object¶: Creates an Operation from a capsule wrapping MlirOperation.

property operation: Operation¶: Returns self (the operation).

property opview: OpView¶

Returns an OpView of this operation.

Note: If the operation has a registered and loaded dialect then this OpView will be concrete wrapper class.

property block: Block¶: Returns the block containing this operation.

property successors: OpSuccessors¶: Returns the list of Operation successors.

replace_uses_of_with(of: Value, with_: Value) → None¶: Replaces uses of the ‘of’ value with the ‘with’ value inside the operation.

_set_invalid() → None¶: Invalidate the operation.

class mlir.dialects.transform.tune.OpView(operation: Operation)¶

class mlir.dialects.transform.tune.OpView(name: str, opRegionSpec: tuple[int, bool], operandSegmentSpecObj: object | None = None, resultSegmentSpecObj: object | None = None, results: collections.abc.Sequence | None = None, operands: collections.abc.Sequence | None = None, attributes: dict[str, Attribute] | None = None, successors: collections.abc.Sequence[Block] | None = None, regions: int | None = None, loc: Location | None = None, ip: object | None = None)

Bases: _OperationBase

property operation: Operation¶

property opview: OpView¶

__str__() → str¶: Returns the assembly form of the operation.

property successors: OpSuccessors¶: Returns the list of Operation successors.

_set_invalid() → None¶: Invalidate the operation.

_ODS_REGIONS: tuple = (0, True)¶

_ODS_OPERAND_SEGMENTS: None = None¶

_ODS_RESULT_SEGMENTS: None = None¶

classmethod build_generic(**kwargs) → Any¶

Builds a specific, generated OpView based on class level attributes.

classmethod parse(**kwargs) → Any¶

parse(cls, source: str, *, source_name: str = ‘’, context: Context | None = None) -> typing.Self

Parses a specific, generated OpView based on class level attributes.

classmethod has_trait(**kwargs) → Any¶

(cls: object, trait_cls: type, context: _mlir.ir.Context | None = None) -> bool

Checks if the operation has a given trait.

class mlir.dialects.transform.tune.Attribute(cast_from_type: Attribute)¶

property _CAPIPtr: object¶: Gets a capsule wrapping the MlirAttribute.

static _CAPICreate(arg: object, /) → Attribute¶: Creates an Attribute from a capsule wrapping MlirAttribute.

static parse(asm: str, context: Context | None = None) → Attribute¶: Parses an attribute from an assembly form. Raises an MLIRError on failure.

property context: Context¶: Context that owns the Attribute.

property type: Type¶: Returns the type of the Attribute.

get_named(arg: str, /) → NamedAttribute¶

Binds a name to the attribute, creating a NamedAttribute.

Parameters:: name – The name to bind to the Attribute.
Returns:: A NamedAttribute with the given name and this attribute.

__eq__(arg: Attribute, /) → bool¶
__eq__(arg: object, /) → bool: Compares attribute with non-attribute object (always returns False).

__hash__() → int¶: Returns the hash value of the attribute.

dump() → None¶: Dumps a debug representation of the object to stderr.

__str__() → str¶: Returns the assembly form of the Attribute.

__repr__() → str¶: Returns a string representation of the attribute.

property typeid: TypeID¶: Returns the TypeID of the attribute.

maybe_downcast() → Attribute¶: Downcasts the attribute to a more specific attribute if possible.

class mlir.dialects.transform.tune.ArrayAttr(cast_from_attr: Attribute)¶

Bases: Attribute

property type: Type¶: Returns the type of the Attribute.

static_typeid: _mlir.ir.TypeID = Ellipsis¶: (arg: object, /) -> _mlir.ir.TypeID

property typeid: TypeID¶: Returns the TypeID of the attribute.

__repr__() → str¶: Returns a string representation of the attribute.

attr_name: str = Ellipsis¶: (arg: object, /) -> str

static get(attributes: collections.abc.Sequence[Attribute], context: Context | None = None) → ArrayAttr¶: Gets a uniqued Array attribute

__getitem__(arg: int, /) → Attribute¶

__len__() → int¶

__iter__() → ArrayAttributeIterator¶

__add__(arg: collections.abc.Sequence[Attribute], /) → ArrayAttr¶

class mlir.dialects.transform.tune.StringAttr(cast_from_attr: Attribute)¶

Bases: Attribute

property type: Type¶: Returns the type of the Attribute.

static_typeid: _mlir.ir.TypeID = Ellipsis¶: (arg: object, /) -> _mlir.ir.TypeID

property typeid: TypeID¶: Returns the TypeID of the attribute.

__repr__() → str¶: Returns a string representation of the attribute.

attr_name: str = Ellipsis¶: (arg: object, /) -> str

static get(value: str, context: Context | None = None) → StringAttr¶
static get(value: bytes, context: Context | None = None) → StringAttr

static get_typed(type: Type, value: str) → StringAttr¶: Gets a uniqued string attribute associated to a type

property value: str¶: Returns the value of the string attribute

property value_bytes: bytes¶: Returns the value of the string attribute as bytes

class mlir.dialects.transform.tune.F64Type(cast_from_type: Type)¶

Bases: FloatType

static_typeid: _mlir.ir.TypeID = Ellipsis¶: (arg: object, /) -> _mlir.ir.TypeID

property typeid: TypeID¶: Returns the TypeID of the Type, or raises ValueError if Type has no TypeID.

__repr__() → str¶: Returns a string representation of the Type.

type_name: str = Ellipsis¶: (arg: object, /) -> str

static get(context: Context | None = None) → F64Type¶: Create a f64 type.

class mlir.dialects.transform.tune.IntegerType(cast_from_type: Type)¶

Bases: Type

static_typeid: _mlir.ir.TypeID = Ellipsis¶: (arg: object, /) -> _mlir.ir.TypeID

property typeid: TypeID¶: Returns the TypeID of the Type, or raises ValueError if Type has no TypeID.

__repr__() → str¶: Returns a string representation of the Type.

type_name: str = Ellipsis¶: (arg: object, /) -> str

class Signedness¶

Bases: enum.Enum

Generic enumeration.

Derive from this class to define new enumerations.

SIGNLESS = 0¶

SIGNED = 1¶

UNSIGNED = 2¶

SIGNLESS: IntegerType¶

SIGNED: IntegerType¶

UNSIGNED: IntegerType¶

static get_signless(width: int, context: Context | None = None) → IntegerType¶: Create a signless integer type

static get_signed(width: int, context: Context | None = None) → IntegerType¶: Create a signed integer type

static get_unsigned(width: int, context: Context | None = None) → IntegerType¶: Create an unsigned integer type

static get(width: int, signedness: IntegerType = IntegerType.Signedness.SIGNLESS, context: Context | None = None) → IntegerType¶: Create an integer type

property signedness: IntegerType¶

property width: int¶: Returns the width of the integer type

property is_signless: bool¶: Returns whether this is a signless integer

property is_signed: bool¶: Returns whether this is a signed integer

property is_unsigned: bool¶: Returns whether this is an unsigned integer

class mlir.dialects.transform.tune.IntegerAttr(cast_from_attr: Attribute)¶

Bases: Attribute

property type: Type¶: Returns the type of the Attribute.

static_typeid: _mlir.ir.TypeID = Ellipsis¶: (arg: object, /) -> _mlir.ir.TypeID

property typeid: TypeID¶: Returns the TypeID of the attribute.

__repr__() → str¶: Returns a string representation of the attribute.

attr_name: str = Ellipsis¶: (arg: object, /) -> str

static get(type: Type, value: object) → IntegerAttr¶: Gets an uniqued integer attribute associated to a type

property value: int¶: Returns the value of the integer attribute

__int__() → int¶: Converts the value of the integer attribute to a Python int

__index__() → int¶: Converts the value of the integer attribute to a Python int

class mlir.dialects.transform.tune.FloatAttr(cast_from_attr: Attribute)¶

Bases: Attribute

property type: Type¶: Returns the type of the Attribute.

static_typeid: _mlir.ir.TypeID = Ellipsis¶: (arg: object, /) -> _mlir.ir.TypeID

property typeid: TypeID¶: Returns the TypeID of the attribute.

__repr__() → str¶: Returns a string representation of the attribute.

attr_name: str = Ellipsis¶: (arg: object, /) -> str

static get(type: Type, value: float, loc: Location | None = None) → FloatAttr¶: Gets an uniqued float point attribute associated to a type

static get_unchecked(type: Type, value: float, context: Context | None = None) → FloatAttr¶: Gets an uniqued float point attribute associated to a type

static get_f32(value: float, context: Context | None = None) → FloatAttr¶: Gets an uniqued float point attribute associated to a f32 type

static get_f64(value: float, context: Context | None = None) → FloatAttr¶: Gets an uniqued float point attribute associated to a f64 type

property value: float¶: Returns the value of the float attribute

__float__() → float¶: Converts the value of the float attribute to a Python float

class mlir.dialects.transform.tune.BoolAttr(cast_from_attr: Attribute)¶

Bases: Attribute

property type: Type¶: Returns the type of the Attribute.

static_typeid: _mlir.ir.TypeID = Ellipsis¶: (arg: object, /) -> _mlir.ir.TypeID

property typeid: TypeID¶: Returns the TypeID of the attribute.

__repr__() → str¶: Returns a string representation of the attribute.

static get(value: bool, context: Context | None = None) → BoolAttr¶: Gets an uniqued bool attribute

property value: bool¶: Returns the value of the bool attribute

__bool__() → bool¶: Converts the value of the bool attribute to a Python bool

Bases: _ods_ir

This op represents a choice over which of its regions is to be used.

When selected_region is provided, the semantics are that this op is to be substituted for by the selected region, meaning the region’s results become the results of this op. Without a provided selected_region, the semantics are that this non-deterministic choice is yet to be resolved – which in terms of the op’s interpreted semantics is a failure.

The selected_region argument is either an IntegerAttr or a param holding an IntegerAttr, which should provide a valid zero-based index with respect to the number of alternatives, i.e. regions.

OPERATION_NAME = 'transform.tune.alternatives'¶

_ODS_REGIONS = (0, False)¶

selected_region_param() → _ods_ir | None¶

name() → _ods_ir¶: Returns the fully qualified name of the operation.

selected_region_attr() → _ods_ir | None¶

results_() → _ods_ir¶

alternatives() → _ods_ir¶

class mlir.dialects.transform.tune.AlternativesOpAdaptor(operands: list[Value], attributes: OpAttributeMap)¶

class mlir.dialects.transform.tune.AlternativesOpAdaptor(operands: list[Value], opview: OpView)

Bases: _ods_ir

OPERATION_NAME = 'transform.tune.alternatives'¶

selected_region_param() → _ods_ir | None¶

name() → _ods_ir¶

selected_region_attr() → _ods_ir | None¶

mlir.dialects.transform.tune.tune_alternatives(results_: Sequence[_ods_ir], name: Any | _ods_ir, num_alternatives: int, *, selected_region_attr: Any | _ods_ir | None = None, selected_region_param: _ods_ir | None = None, loc: _ods_ir | None = None, ip: _ods_ir | None = None) → _ods_ir | _ods_ir | AlternativesOp¶

Bases: _ods_ir

Provides a representation for “tunables” within schedules.

Each op represents a single tunable, which has a name and a set of valid options described by an attribute. Without a specified selected option, this op represents a non-deterministic choice that has yet to be resolved – as such, the interpreter runtime semantics is to raise a failure.

The non-deterministic choice is resolved through providing a selected attribute. When provided, the interpreter runtime semantics are to return the selected attribute as a param through the op’s result.

In case the options attribute is an ArrayAttr, the verifier checks that the provided selected attribute occurs in options.

OPERATION_NAME = 'transform.tune.knob'¶

_ODS_REGIONS = (0, True)¶

name() → _ods_ir¶: Returns the fully qualified name of the operation.

options() → _ods_ir¶

selected() → _ods_ir | None¶

result() → _ods_ir¶: Shortcut to get an op result if it has only one (throws an error otherwise).

class mlir.dialects.transform.tune.KnobOpAdaptor(operands: list[Value], attributes: OpAttributeMap)¶

class mlir.dialects.transform.tune.KnobOpAdaptor(operands: list[Value], opview: OpView)

Bases: _ods_ir

OPERATION_NAME = 'transform.tune.knob'¶

name() → _ods_ir¶

options() → _ods_ir¶

selected() → _ods_ir | None¶

mlir.dialects.transform.tune.tune_knob(result: _ods_ir, name: Any | _ods_ir, options: Any | _ods_ir, *, selected: Any | _ods_ir | None = None, loc: _ods_ir | None = None, ip: _ods_ir | None = None) → _ods_ir¶

Bases: KnobOp

Provides a representation for “tunables” within schedules.

In case the options attribute is an ArrayAttr, the verifier checks that the provided selected attribute occurs in options.

mlir.dialects.transform.tune.knob(result: mlir.ir.Type, name: mlir.ir.StringAttr | str, options: mlir.ir.ArrayAttr | Sequence[mlir.ir.Attribute | bool | int | float | str] | mlir.ir.Attribute, *, selected: mlir.ir.Attribute | bool | int | float | str | None = None, loc=None, ip=None)¶

Bases: AlternativesOp

This op represents a choice over which of its regions is to be used.

The selected_region argument is either an IntegerAttr or a param holding an IntegerAttr, which should provide a valid zero-based index with respect to the number of alternatives, i.e. regions.

mlir.dialects.transform.tune.alternatives(results: Sequence[mlir.ir.Type], name: mlir.ir.StringAttr | str, num_alternatives: int, *, selected_region: int | mlir.ir.IntegerAttr | mlir.ir.Value | mlir.ir.Operation | mlir.ir.OpView | None = None, loc=None, ip=None)¶