MLIR

Multi-Level IR Compiler Framework

'func' Dialect

This dialect provides documentation for operations within the Func dialect.

This dialect contains operations surrounding high order function abstractions, such as calls.

Please post an RFC on the forum before adding or changing any operation in this dialect.

Operations 

source

func.call_indirect (func::CallIndirectOp) 

Indirect call operation

Syntax:

operation ::= `func.call_indirect` $callee `(` $callee_operands `)` attr-dict `:` type($callee)

The func.call_indirect operation represents an indirect call to a value of function type. The operands and result types of the call must match the specified function type.

Function values can be created with the func.constant operation.

Example:

%func = func.constant @my_func : (tensor<16xf32>, tensor<16xf32>) -> tensor<16xf32>
%result = func.call_indirect %func(%0, %1) : (tensor<16xf32>, tensor<16xf32>) -> tensor<16xf32>

Interfaces: CallOpInterface

Operands: 

OperandDescription
calleefunction type
callee_operandsvariadic of any type

Results: 

ResultDescription
resultsvariadic of any type

func.call (func::CallOp) 

Call operation

Syntax:

operation ::= `func.call` $callee `(` $operands `)` attr-dict `:` functional-type($operands, results)

The func.call operation represents a direct call to a function that is within the same symbol scope as the call. The operands and result types of the call must match the specified function type. The callee is encoded as a symbol reference attribute named “callee”.

Example:

%2 = func.call @my_add(%0, %1) : (f32, f32) -> f32

Traits: MemRefsNormalizable

Interfaces: CallOpInterface, SymbolUserOpInterface

Attributes: 

AttributeMLIR TypeDescription
callee::mlir::FlatSymbolRefAttrflat symbol reference attribute

Operands: 

OperandDescription
operandsvariadic of any type

Results: 

ResultDescription
«unnamed»variadic of any type

func.constant (func::ConstantOp) 

Constant

Syntax:

operation ::= `func.constant` attr-dict $value `:` type(results)

The func.constant operation produces an SSA value from a symbol reference to a func.func operation

Example:

// Reference to function @myfn.
%2 = func.constant @myfn : (tensor<16xf32>, f32) -> tensor<16xf32>

// Equivalent generic forms
%2 = "func.constant"() { value = @myfn } : () -> ((tensor<16xf32>, f32) -> tensor<16xf32>)

MLIR does not allow direct references to functions in SSA operands because the compiler is multithreaded, and disallowing SSA values to directly reference a function simplifies this ( rationale).

Traits: AlwaysSpeculatableImplTrait, ConstantLike

Interfaces: ConditionallySpeculatable, NoMemoryEffect (MemoryEffectOpInterface), OpAsmOpInterface

Effects: MemoryEffects::Effect{}

Attributes: 

AttributeMLIR TypeDescription
value::mlir::FlatSymbolRefAttrflat symbol reference attribute

Results: 

ResultDescription
«unnamed»any type

func.func (func::FuncOp) 

An operation with a name containing a single SSACFG region

Operations within the function cannot implicitly capture values defined outside of the function, i.e. Functions are IsolatedFromAbove. All external references must use function arguments or attributes that establish a symbolic connection (e.g. symbols referenced by name via a string attribute like SymbolRefAttr). An external function declaration (used when referring to a function declared in some other module) has no body. While the MLIR textual form provides a nice inline syntax for function arguments, they are internally represented as “block arguments” to the first block in the region.

Only dialect attribute names may be specified in the attribute dictionaries for function arguments, results, or the function itself.

Example:

// External function definitions.
func.func private @abort()
func.func private @scribble(i32, i64, memref<? x 128 x f32, #layout_map0>) -> f64

// A function that returns its argument twice:
func.func @count(%x: i64) -> (i64, i64)
  attributes {fruit: "banana"} {
  return %x, %x: i64, i64
}

// A function with an argument attribute
func.func private @example_fn_arg(%x: i32 {swift.self = unit})

// A function with a result attribute
func.func private @example_fn_result() -> (f64 {dialectName.attrName = 0 : i64})

// A function with an attribute
func.func private @example_fn_attr() attributes {dialectName.attrName = false}

Traits: AffineScope, AutomaticAllocationScope, IsolatedFromAbove

Interfaces: CallableOpInterface, FunctionOpInterface, OpAsmOpInterface, Symbol

Attributes: 

AttributeMLIR TypeDescription
sym_name::mlir::StringAttrstring attribute
function_type::mlir::TypeAttrtype attribute of function type
sym_visibility::mlir::StringAttrstring attribute
arg_attrs::mlir::ArrayAttrArray of dictionary attributes
res_attrs::mlir::ArrayAttrArray of dictionary attributes

func.return (func::ReturnOp) 

Function return operation

Syntax:

operation ::= `func.return` attr-dict ($operands^ `:` type($operands))?

The func.return operation represents a return operation within a function. The operation takes variable number of operands and produces no results. The operand number and types must match the signature of the function that contains the operation.

Example:

func.func @foo() : (i32, f8) {
  ...
  return %0, %1 : i32, f8
}

Traits: AlwaysSpeculatableImplTrait, HasParent<FuncOp>, MemRefsNormalizable, ReturnLike, Terminator

Interfaces: ConditionallySpeculatable, NoMemoryEffect (MemoryEffectOpInterface), RegionBranchTerminatorOpInterface

Effects: MemoryEffects::Effect{}

Operands: 

OperandDescription
operandsvariadic of any type