MLIR

Multi-Level IR Compiler Framework

Table-driven Declarative Rewrite Rule (DRR)

In addition to subclassing the mlir::RewritePattern C++ class, MLIR also supports defining rewrite rules in a declarative manner. Similar to Op Definition Specification (ODS), this is achieved via TableGen , which is a language to maintain records of domain-specific information. The rewrite rules are specified concisely in a TableGen record, which will be expanded into an equivalent mlir::RewritePattern subclass at compiler build time.

This manual explains in detail all of the available mechanisms for defining rewrite rules in such a declarative manner. It aims to be a specification instead of a tutorial. Please refer to Quickstart tutorial to adding MLIR graph rewrite for the latter.

Given that declarative rewrite rules depend on op definition specification, this manual assumes knowledge of the ODS doc.

Benefits

Compared to the hand-written C++ classes, this declarative approach has several benefits, including but not limited to:

  • Being declarative: The pattern creator just needs to state the rewrite pattern declaratively, without worrying about the concrete C++ methods to call.
  • Removing boilerplate and showing the very essence of the rewrite: mlir::RewritePattern is already good at hiding boilerplate for defining a rewrite rule. But we still need to write the class and function structures required by the C++ programming language, inspect ops for matching, and call op build() methods for constructing. These statements are typically quite simple and similar, so they can be further condensed with auto-generation. Because we reduce the boilerplate to the bare minimum, the declarative rewrite rule will just contain the very essence of the rewrite. This makes it very easy to understand the pattern.

Strengths and Limitations

The declarative rewrite rule is operation-based: it describes a rule to match against a directed acyclic graph (DAG) of operations and generate DAGs of operations. This gives DRR both its strengths and limitations: it is good at expressing op to op conversions, but not that well suited for, say, converting an op into a loop nest.

Per the current implementation, DRR does not have good support for the following features:

  • Matching and generating ops with regions.
  • Matching and generating ops with block arguments.
  • Matching multi-result ops in nested patterns.
  • Matching and generating variadic operand/result ops in nested patterns.
  • Packing and unpacking variadic operands/results during generation.
  • NativeCodeCall returning more than one results.

Rule Definition

The core construct for defining a rewrite rule is defined in OpBase.td as

class Pattern<
    dag sourcePattern, list<dag> resultPatterns,
    list<dag> additionalConstraints = [],
    dag benefitsAdded = (addBenefit 0)>;

A declarative rewrite rule contains two main components:

  • A source pattern, which is used for matching a DAG of operations.
  • One or more result patterns, which are used for generating DAGs of operations to replace the matched DAG of operations.

We allow multiple result patterns to support multi-result ops and auxiliary ops , but frequently we just want to convert one DAG of operations to another DAG of operations. There is a handy wrapper of Pattern, Pat, which takes a single result pattern:

class Pat<
    dag sourcePattern, dag resultPattern,
    list<dag> additionalConstraints = [],
    dag benefitsAdded = (addBenefit 0)> :
  Pattern<sourcePattern, [resultPattern], additionalConstraints, benefitAdded>;

Each pattern is specified as a TableGen dag object with the syntax of (operator arg0, arg1, ...).

operator is typically an MLIR op, but it can also be other directives . argN is for matching (if used in source pattern) or generating (if used in result pattern) the N-th argument for operator. If the operator is some MLIR operation, it means the N-th argument as specified in the arguments list of the op’s definition. Therefore, we say op argument specification in pattern is position-based: the position where they appear matters.

argN can be a dag object itself, thus we can have nested dag tree to model the def-use relationship between ops.

Source pattern

The source pattern is for matching a DAG of operations. Arguments in the dag object are intended to capture the op arguments. They can also be used to further limit the match criteria. The capturing is done by specifying a symbol starting with the $ sign, while further constraints are introduced by specifying a TypeConstraint (for an operand) or a AttrConstraint (for an attribute).

Binding op arguments and limiting the match

For example,

def AOp : Op<"a_op"> {
    let arguments = (ins
      AnyType:$a_input,
      AnyAttr:$a_attr
    );

    let results = (outs
      AnyType:$a_output
    );
}

def : Pat<(AOp $input, F32Attr:$attr), ...>;

In the above, we are matching an AOp whose $input can be anything valid as defined by the op and whose $attr must be a float attribute. If the match succeeds, we bind the $input symbol to the op’s only input ($a_input) and $attr to the only attribute ($a_attr); we can reference them using $input and $attr in result patterns and additional constraints.

The pattern is position-based: the symbol names used for capturing here do not need to match with the op definition as shown in the above example. As another example, the pattern can be written as def : Pat<(AOp $a, F32Attr:$b), ...>; and use $a and $b to refer to the captured input and attribute. But using the ODS name directly in the pattern is also allowed.

Also note that we only need to add TypeConstraint or AttributeConstraint when we need to further limit the match criteria. If all valid cases to the op are acceptable, then we can leave the constraint unspecified.

$_ is a special symbol to mean ignore capturing an argument. For example, def : Pat<(AOp $_, $b), ...> means only $b is interesting to capture and will be referenced later in result patterns. It’s still possible to place additional constraints even if the symbol is not to be captured; for such case, you can simply use just the TypeConstraint or AttributeConstraint without a bound symbol, for example, def : Pat<(AOp $a, F32Attr), ...>.

Matching DAG of operations

To match an DAG of ops, use nested dag objects:


def BOp : Op<"b_op"> {
    let arguments = (ins);

    let results = (outs
      AnyType:$b_output
    );
}


def : Pat<(AOp (BOp), $attr), ...>;

The above pattern matches an AOp whose only operand is generated by a BOp, that is, the following MLIR code:

%0 = "b_op"() : () -> (...)
%1 = "a_op"(%0) {attr: ...} : () -> (...)

Binding op results

To bind a symbol to the results of a matched op for later reference, attach the symbol to the op itself:

def : Pat<(AOp (BOp:$b_result), $attr), ...>;

The above will bind $b_result to the matched BOp's result. (There are more details regarding multi-result ops, which is covered later .)

Result pattern

The result pattern is for generating a DAG of operations. Arguments in the dag object are intended to reference values captured in the source pattern and potentially apply transformations.

Referencing bound symbols

For example,

def COp : Op<"c_op"> {
    let arguments = (ins
      AnyType:$c_input,
      AnyAttr:$c_attr
    );

    let results = (outs
      AnyType:$c_output
    );
}

def : Pat<(AOp $input, $attr), (COp $input, $attr)>;

In the above, AOp's only operand and attribute are bound to $input and $attr, respectively. We then reference them in the result pattern for generating the COp by passing them in as arguments to COp's build() method.

We can also reference symbols bound to matched op’s results:

def : Pat<(AOp (BOp:$b_result) $attr), (COp $b_result $attr)>;

In the above, we are using BOp's result for building COp.

Building operations

Given that COp was specified with table-driven op definition, there will be several build() methods generated for it. One of them has aggregated parameters for result types, operands, and attributes in the signature: void COp::build(..., ArrayRef<Type> resultTypes, Array<Value> operands, ArrayRef<NamedAttribute> attr). The pattern in the above calls this build() method for constructing the COp.

In general, arguments in the result pattern will be passed directly to the build() method to leverage the auto-generated build() method, list them in the pattern by following the exact same order as the ODS arguments definition. Otherwise, a custom build() method that matches the argument list is required.

Right now all ODS-generated build() methods require specifying the result type(s), unless the op has known traits like SameOperandsAndResultType that we can use to auto-generate a build() method with result type deduction. When generating an op to replace the result of the matched root op, we can use the matched root op’s result type when calling the ODS-generated builder. Otherwise (e.g., generating an auxiliary op or generating an op with a nested result pattern), DRR will not be able to deduce the result type(s). The pattern author will need to define a custom builder that has result type deduction ability via OpBuilder in ODS. For example, in the following pattern

def : Pat<(AOp $input, $attr), (COp (AOp $input, $attr) $attr)>;

AOp is generated via a nested result pattern; DRR won’t be able to deduce the result type for it. A custom builder for AOp should be defined and it should deduce the result type by itself. The builder should have the separate parameter for each operand and attribute and deduce the result type internally by itself. For example, for the above AOp, a possible builder is:


void AOp::build(Builder *builder, OperationState &state,
                Value input, Attribute attr) {
  state.addOperands({input});
  state.addAttribute("a_attr", attr);
  Type type = ...; // Deduce result type here
  state.addTypes({type});
}

Failing to define such a builder will result in an error at C++ compilation time saying the call to AOp::build() cannot be resolved because of the number of parameters mismatch.

Generating DAG of operations

dag objects can be nested to generate a DAG of operations:

def : Pat<(AOp $input, $attr), (COp (BOp), $attr)>;

In the above, we generate a BOp, and then use its result to generate the COp to replace the matched AOp.

Binding op results

In the result pattern, we can bind to the result(s) of a newly built op by attaching symbols to the op. (But we cannot bind to op arguments given that they are referencing previously bound symbols.) This is useful for reusing newly created results where suitable. For example,

def DOp : Op<"d_op"> {
    let arguments = (ins
      AnyType:$d_input1,
      AnyType:$d_input2,
    );

    let results = (outs
      AnyType:$d_output
    );
}

def : Pat<(AOp $input, $ignored_attr), (DOp (BOp:$b_result) $b_result)>;

In this pattern, an AOp is matched and replaced with a DOp whose two operands are from the result of a single BOp. This is only possible by binding the result of the BOp to a name and reuse it for the second operand of the DOp

NativeCodeCall: transforming the generated op

Sometimes the captured arguments are not exactly what we want so they cannot be directly fed in as arguments to build the new op. For such cases, we can apply transformations on the arguments by calling into C++ helper functions. This is achieved by NativeCodeCall.

For example, if we want to capture some op’s attributes and group them as an array attribute to construct a new op:


def TwoAttrOp : Op<"two_attr_op"> {
    let arguments = (ins
      AnyAttr:$op_attr1,
      AnyAttr:$op_attr2
    );

    let results = (outs
      AnyType:$op_output
    );
}

def OneAttrOp : Op<"one_attr_op"> {
    let arguments = (ins
      ArrayAttr:$op_attr
    );

    let results = (outs
      AnyType:$op_output
    );
}

We can write a C++ helper function:

Attribute createArrayAttr(Builder &builder, Attribute a, Attribute b) {
  return builder.getArrayAttr({a, b});
}

And then write the pattern as:

def createArrayAttr : NativeCodeCall<"createArrayAttr($_builder, $0, $1)">;

def : Pat<(TwoAttrOp $attr1, $attr2),
          (OneAttrOp (createArrayAttr $attr1, $attr2))>;

And make sure the generated C++ code from the above pattern has access to the definition of the C++ helper function.

In the above example, we are using a string to specialize the NativeCodeCall template. The string can be an arbitrary C++ expression that evaluates into some C++ object expected at the NativeCodeCall site (here it would be expecting an array attribute). Typically the string should be a function call.

Note that currently NativeCodeCall must return no more than one value or attribute. This might change in the future.

NativeCodeCall placeholders

In NativeCodeCall, we can use placeholders like $_builder, $N. The former is called special placeholder, while the latter is called positional placeholder.

NativeCodeCall right now only supports two special placeholders: $_builder and $_self:

  • $_builder will be replaced by the current mlir::PatternRewriter.
  • $_self will be replaced with the entity NativeCodeCall is attached to.

We have seen how $_builder can be used in the above; it allows us to pass a mlir::Builder (mlir::PatternRewriter is a subclass of mlir::OpBuilder, which is a subclass of mlir::Builder) to the C++ helper function to use the handy methods on mlir::Builder.

$_self is useful when we want to write something in the form of NativeCodeCall<"...">:$symbol. For example, if we want to reverse the previous example and decompose the array attribute into two attributes:

class getNthAttr<int n> : NativeCodeCall<"$_self[" # n # "]">;

def : Pat<(OneAttrOp $attr),
          (TwoAttrOp (getNthAttr<0>:$attr), (getNthAttr<1>:$attr)>;

In the above, $_self is substituted by the attribute bound by $attr, which is OneAttrOp's array attribute.

Positional placeholders will be substituted by the dag object parameters at the NativeCodeCall use site. For example, if we define SomeCall : NativeCodeCall<"someFn($1, $2, $0)"> and use it like (SomeCall $in0, $in1, $in2), then this will be translated into C++ call someFn($in1, $in2, $in0).

Customizing entire op building

NativeCodeCall is not only limited to transforming arguments for building an op; it can be also used to specify how to build an op entirely. An example:

If we have a C++ function for building an op:

Operation *createMyOp(OpBuilder builder, Value input, Attribute attr);

We can wrap it up and invoke it like:

def createMyOp : NativeCodeCall<"createMyOp($_builder, $0, $1)">;

def : Pat<(... $input, $attr), (createMyOp $input, $attr)>;

Supporting auxiliary ops

A declarative rewrite rule supports multiple result patterns. One of the purposes is to allow generating auxiliary ops. Auxiliary ops are operations used for building the replacement ops; but they are not directly used for replacement themselves.

For the case of uni-result ops, if there are multiple result patterns, only the value generated from the last result pattern will be used to replace the matched root op’s result; all other result patterns will be considered as generating auxiliary ops.

Normally we want to specify ops as nested dag objects if their def-use relationship can be expressed in the way that an op’s result can feed as the argument to consuming op. But that is not always possible. For example, if we want to allocate memory and store some computation (in pseudocode):

%dst = addi %lhs, %rhs

into

%shape = shape %lhs
%mem = alloc %shape
%sum = addi %lhs, %rhs
store %mem, %sum
%dst = load %mem

We cannot fit in with just one result pattern given store does not return a value. Instead we can use multiple result patterns:

def : Pattern<(AddIOp $lhs, $rhs),
              [(StoreOp (AllocOp:$mem (ShapeOp $lhs)), (AddIOp $lhs, $rhs)),
               (LoadOp $mem)];

In the above we use the first result pattern to generate the first four ops, and use the last pattern to generate the last op, which is used to replace the matched op.

Supporting multi-result ops

Multi-result ops bring extra complexity to declarative rewrite rules. We use TableGen dag objects to represent ops in patterns; there is no native way to indicate that an op generates multiple results. The approach adopted is based on naming convention: a __N suffix is added to a symbol to indicate the N-th result.

__N suffix

The __N suffix is specifying the N-th result as a whole (which can be variadic ). For example, we can bind a symbol to some multi-result op and reference a specific result later:

def ThreeResultOp : Op<"three_result_op"> {
    let arguments = (ins ...);

    let results = (outs
      AnyTensor:$op_output1,
      AnyTensor:$op_output2,
      AnyTensor:$op_output3
    );
}

def : Pattern<(ThreeResultOp:$results ...),
              [(... $results__0), ..., (... $results__2), ...]>;

In the above pattern we bind $results to all the results generated by ThreeResultOp and references its $input1 and $input3 later in the result patterns.

We can also bind a symbol and reference one of its specific result at the same time, which is typically useful when generating multi-result ops:

// TwoResultOp has similar definition as ThreeResultOp, but only has two
// results.

def : Pattern<(TwoResultOp ...),
              [(ThreeResultOp:$results__2, ...),
               (replaceWithValue $results__0)]>;

In the above, we created a ThreeResultOp and bind results to its results, and uses its last result ($output3) and first result ($output1) to replace the TwoResultOp's two results, respectively.

Replacing multi-result ops

The above example also shows how to replace a matched multi-result op.

To replace a N-result op, the result patterns must generate at least N declared values (see Declared vs. actual value for definition). If there are more than N declared values generated, only the last N declared values will be used to replace the matched op. Note that because of the existence of multi-result op, one result pattern may generate multiple declared values. So it means we do not necessarily need N result patterns to replace an N-result op. For example, to replace an op with three results, you can have

// ThreeResultOp/TwoResultOp/OneResultOp generates three/two/one result(s),
// respectively.

// Replace each result with a result generated from an individual op.
def : Pattern<(ThreeResultOp ...),
              [(OneResultOp ...), (OneResultOp ...), (OneResultOp ...)]>;

// Replace the first two results with two results generated from the same op.
def : Pattern<(ThreeResultOp ...),
              [(TwoResultOp ...), (OneResultOp ...)]>;

// Replace all three results with three results generated from the same op.
def : Pat<(ThreeResultOp ...), (ThreeResultOp ...)>;

def : Pattern<(ThreeResultOp ...),
              [(AuxiliaryOp ...), (ThreeResultOp ...)]>;

But using a single op to serve as both auxiliary op and replacement op is forbidden, i.e., the following is not allowed because that the first TwoResultOp generates two results but only the second result is used for replacing the matched op’s result:

def : Pattern<(ThreeResultOp ...),
              [(TwoResultOp ...), (TwoResultOp ...)]>;

Supporting variadic ops

Declared vs. actual value

Before going into details on variadic op support, we need to define a few terms regarding an op’s values.

  • Value: either an operand or a result
  • Declared operand/result/value: an operand/result/value statically declared in ODS of the op
  • Actual operand/result/value: an operand/result/value of an op instance at runtime

The above terms are needed because ops can have multiple results, and some of the results can also be variadic. For example,

def MultiVariadicOp : Op<"multi_variadic_op"> {
    let arguments = (ins
      AnyTensor:$input1,
      Variadic<AnyTensor>:$input2,
      AnyTensor:$input3
    );

    let results = (outs
      AnyTensor:$output1,
      Variadic<AnyTensor>:$output2,
      AnyTensor:$output3
    );
}

We say the above op has 3 declared operands and 3 declared results. But at runtime, an instance can have 3 values corresponding to $input2 and 2 values correspond to $output2; we say it has 5 actual operands and 4 actual results. A variadic operand/result is a considered as a declared value that can correspond to multiple actual values.

[TODO]

Supplying additional constraints

Constraints can be placed on op arguments when matching. But sometimes we need to also place constraints on the matched op’s results or sometimes need to limit the matching with some constraints that cover both the arguments and the results. The third parameter to Pattern (and Pat) is for this purpose.

For example, we can write

def HasNoUseOf: Constraint<CPred<"$_self.use_empty()">, "has no use">;

def HasSameElementType : Constraint<
    CPred<"$0.cast<ShapedType>().getElementType() == "
          "$1.cast<ShapedType>().getElementType()">,
    "has same element type">;

def : Pattern<(TwoResultOp:$results $input),
              [(...), (...)],
              [(F32Tensor:$results__0), (HasNoUseOf:$results__1),
               (HasSameElementShape $results__0, $input)]>;

You can

  • Use normal TypeConstraints on previous bound symbols (the first result of TwoResultOp must be a float tensor);
  • Define new Constraint for previous bound symbols (the second result of TwoResultOp must has no use);
  • Apply constraints on multiple bound symbols ($input and TwoResultOp's first result must have the same element type).

Adjusting benefits

The benefit of a Pattern is an integer value indicating the benefit of matching the pattern. It determines the priorities of patterns inside the pattern rewrite driver. A pattern with a higher benefit is applied before one with a lower benefit.

In DRR, a rule is set to have a benefit of the number of ops in the source pattern. This is based on the heuristics and assumptions that:

  • Larger matches are more beneficial than smaller ones.
  • If a smaller one is applied first the larger one may not apply anymore.

The fourth parameter to Pattern (and Pat) allows to manually tweak a pattern’s benefit. Just supply (addBenefit N) to add N to the benefit value.

Rewrite directives

location

By default the C++ pattern expanded from a DRR pattern uses the fused location of all source ops as the location for all generated ops. This is not always the best location mapping relationship. For such cases, DRR provides the location directive to provide finer control.

location is of the following syntax:

(location $symbol0, $symbol1, ...)

where all $symbol should be bound previously in the pattern.

location must be used as the last argument to an op creation. For example,

def : Pat<(LocSrc1Op:$src1 (LocSrc2Op:$src2 ...),
          (LocDst1Op (LocDst2Op ..., (location $src2)))>;

In the above pattern, the generated LocDst2Op will use the matched location of LocSrc2Op while the root LocDst1Op node will still se the fused location of all source Ops.

replaceWithValue

The replaceWithValue directive is used to eliminate a matched op by replacing all of it uses with a captured value. It is of the following syntax:

(replaceWithValue $symbol)

where $symbol should be a symbol bound previously in the pattern.

For example,

def : Pat<(Foo $input), (replaceWithValue $input)>;

The above pattern removes the Foo and replaces all uses of Foo with $input.

Debugging Tips

Run mlir-tblgen to see the generated content

TableGen syntax sometimes can be obscure; reading the generated content can be a very helpful way to understand and debug issues. To build mlir-tblgen, run cmake --build . --target mlir-tblgen in your build directory and find the mlir-tblgen binary in the bin/ subdirectory. All the supported generators can be found via mlir-tblgen --help.

To see the generated code, invoke mlir-tblgen with a specific generator by providing include paths via -I. For example,

# To see all the C++ pattern rewrite classes
mlir-tblgen --gen-rewriters -I /path/to/mlir/include /path/to/input/td/file

Compilation error: no matching member function for call to ‘build’

This is because DRR is failing to call a build() method with result type deduction ability. See building operations for more details.