Data Layout Modeling

Data layout information allows the compiler to answer questions related to how a value of a particular type is stored in memory. For example, the size of a value or its address alignment requirements. It enables, among others, the generation of various linear memory addressing schemes for containers of abstract types and deeper reasoning about vectors.

The data layout subsystem is designed to scale to MLIR’s open type and operation system. At the top level, it consists of:

attribute interfaces that can be implemented by concrete data layout specifications;
type interfaces that should be implemented by types subject to data layout;
operation interfaces that must be implemented by operations that can serve as data layout scopes (e.g., modules);
and dialect interfaces for data layout properties unrelated to specific types.

Built-in types are handled specially to decrease the overall query cost. Similarly, built-in ModuleOp supports data layouts without going through the interface.

Usage ¶

Scoping ¶

Following MLIR’s nested structure, data layout properties are scoped to regions belonging to either operations that implement the DataLayoutOpInterface or ModuleOp operations. Such scoping operations partially control the data layout properties and may have attributes that affect them, typically organized in a data layout specification.

Types may have a different data layout in different scopes, including scopes that are nested in other scopes such as modules contained in other modules. At the same time, within the given scope excluding any nested scope, a given type has fixed data layout properties. Types are also expected to have a default, “natural” data layout in case they are used outside of any operation that provides data layout scope for them. This ensures that data layout queries always have a valid result.

Compatibility and Transformations ¶

The information necessary to compute layout properties can be combined from nested scopes. For example, an outer scope can define layout properties for a subset of types while inner scopes define them for a disjoint subset, or scopes can progressively relax alignment requirements on a type. This mechanism is supported by the notion of data layout compatibility: the layout defined in a nested scope is expected to be compatible with that of the outer scope. MLIR does not prescribe what compatibility means for particular ops and types but provides hooks for them to provide target- and type-specific checks. For example, one may want to only allow relaxation of alignment constraints (i.e., smaller alignment) in nested modules or, alternatively, one may require nested modules to fully redefine all constraints of the outer scope.

Data layout compatibility is also relevant during IR transformation. Any transformation that affects the data layout scoping operation is expected to maintain data layout compatibility. It is under responsibility of the transformation to ensure it is indeed the case.

Queries ¶

Data layout property queries can be performed on the special object – DataLayout – which can be created for the given scoping operation. These objects allow one to interface with the data layout infrastructure and query properties of given types in the scope of the object. The signature of DataLayout class is as follows.

class DataLayout {
public:
  explicit DataLayout(DataLayoutOpInterface scope);

  llvm::TypeSize getTypeSize(Type type) const;
  llvm::TypeSize getTypeSizeInBits(Type type) const;
  uint64_t getTypeABIAlignment(Type type) const;
  uint64_t getTypePreferredAlignment(Type type) const;
  std::optional<uint64_t> getTypeIndexBitwidth(Type type) const;
};

The user can construct the DataLayout object for the scope of interest. Since the data layout properties are fixed in the scope, they will be computed only once upon first request and cached for further use. Therefore, DataLayout(op.getParentOfType<DataLayoutOpInterface>()).getTypeSize(type) is considered an anti-pattern since it discards the cache after use. Because of caching, a DataLayout object returns valid results as long as the data layout properties of enclosing scopes remain the same, that is, as long as none of the ancestor operations are modified in a way that affects data layout. After such a modification, the user is expected to create a fresh DataLayout object. To aid with this, DataLayout asserts that the scope remains identical if MLIR is compiled with assertions enabled.

Custom Implementations ¶

Extensibility of the data layout modeling is provided through a set of MLIR Interfaces.

Data Layout Specifications ¶

Data layout specification is an attribute that is conceptually a collection of key-value pairs called data layout specification entries. Data layout specification attributes implement the DataLayoutSpecInterface, described below. Each entry is itself an attribute that implements the DataLayoutEntryInterface. Entries have a key, either a Type or a StringAttr, and a value. Keys are used to associate entries with specific types or dialects: when handling a data layout properties request, a type or a dialect can only see the specification entries relevant to them and must go through the supplied DataLayout object for any recursive query. This supports and enforces better composability because types cannot (and should not) understand layout details of other types. Entry values are arbitrary attributes, specific to the type.

For example, a data layout specification may be an actual list of pairs with simple custom syntax resembling the following:

#my_dialect.layout_spec<
  #my_dialect.layout_entry<!my_dialect.type, size=42>,
  #my_dialect.layout_entry<"my_dialect.endianness", "little">,
  #my_dialect.layout_entry<!my_dialect.vector, prefer_large_alignment>>

The exact details of the specification and entry attributes, as well as their syntax, are up to implementations.

We use the notion of type class throughout the data layout subsystem. It corresponds to the C++ class of the given type, e.g., IntegerType for built-in integers. MLIR does not have a mechanism to represent type classes in the IR. Instead, data layout entries contain specific instances of a type class, for example, IntegerType{signedness=signless, bitwidth=8} (or i8 in the IR) or IntegerType{signedness=unsigned, bitwidth=32} (or ui32 in the IR). When handling a data layout property query, a type class will be supplied with all entries with keys belonging to this type class. For example, IntegerType will see the entries for i8, si16 and ui32, but will not see those for f32 or memref<?xi32> (neither will MemRefType see the entry for i32). This allows for type-specific “interpolation” behavior where a type class can compute data layout properties of any specific type instance given properties of other instances. Using integers as an example again, their alignment could be computed by taking that of the closest from above integer type with power-of-two bitwidth.

DataLayoutEntryInterface (`DataLayoutEntryInterface`) ¶

Attribute interface describing an entry in a data layout specification.

A data layout specification entry is a key-value pair. Its key is either a type, when the entry is related to a type or a class of types, or an identifier, when it is not. DataLayoutEntryKey is an alias allowing one to use both key types. Its value is an arbitrary attribute that is interpreted either by the type for type keys or by the dialect containing the identifier for identifier keys. The interface provides a hook that can be used by specific implementations to delegate the verification of attribute fitness for a particular key to the relevant type or dialect.

Methods: ¶

`getKey` ¶

::mlir::DataLayoutEntryKey getKey();

Returns the key of the this layout entry. NOTE: This method must be implemented by the user.

`getValue` ¶

::mlir::Attribute getValue();

Returns the value of this layout entry. NOTE: This method must be implemented by the user.

`verifyEntry` ¶

::mlir::LogicalResult verifyEntry(::mlir::Location loc);

Checks that the entry is well-formed, reports errors at the provided location. NOTE: This method must be implemented by the user.

DataLayoutSpecInterface (`DataLayoutSpecInterface`) ¶

Attribute interface describing a data layout specification.

A data layout specification is seen as a sequence of entries, each of which is an attribute implementing the data layout entry interface. It assumes a contiguous underlying storage for entries. The interface provides a hook for implementations to verify the well-formedness of the specification, with a default implementation that verifies the absence of entries with duplicate keys and the well-formedness of each individual entry before dispatching to the type or dialect the entry is associated with.

Data layout specifications may need to be combined in case they appear on nested operations subject to layout, or to ensure the validity of layout modification. Concrete specification attributes must implement the corresponding hook.

Methods: ¶

`combineWith` ¶

::mlir::DataLayoutSpecInterface combineWith(::llvm::ArrayRef<::mlir::DataLayoutSpecInterface> specs);

Combines the current layout with the given list of layouts, provided from the outermost (oldest) to the innermost (newest). Returns null on failure. NOTE: This method must be implemented by the user.

`getEntries` ¶

::mlir::DataLayoutEntryListRef getEntries();

Returns the list of layout entries. NOTE: This method must be implemented by the user.

`getEndiannessIdentifier` ¶

::mlir::StringAttr getEndiannessIdentifier(::mlir::MLIRContext *context);

Returns the endianness identifier. NOTE: This method must be implemented by the user.

`getAllocaMemorySpaceIdentifier` ¶

::mlir::StringAttr getAllocaMemorySpaceIdentifier(::mlir::MLIRContext *context);

Returns the alloca memory space identifier. NOTE: This method must be implemented by the user.

`getProgramMemorySpaceIdentifier` ¶

::mlir::StringAttr getProgramMemorySpaceIdentifier(::mlir::MLIRContext *context);

Returns the program memory space identifier. NOTE: This method must be implemented by the user.

`getGlobalMemorySpaceIdentifier` ¶

::mlir::StringAttr getGlobalMemorySpaceIdentifier(::mlir::MLIRContext *context);

Returns the global memory space identifier. NOTE: This method must be implemented by the user.

`getStackAlignmentIdentifier` ¶

::mlir::StringAttr getStackAlignmentIdentifier(::mlir::MLIRContext *context);

Returns the stack alignment identifier. NOTE: This method must be implemented by the user.

`getSpecForType` ¶

::mlir::DataLayoutEntryList getSpecForType(::mlir::TypeID type);

Returns a copy of the entries related to a specific type class regardles of type parameters. NOTE: This method must be implemented by the user.

`getSpecForIdentifier` ¶

::mlir::DataLayoutEntryInterface getSpecForIdentifier(::mlir::StringAttr identifier);

Returns the entry related to the given identifier, if present. NOTE: This method must be implemented by the user.

`verifySpec` ¶

::mlir::LogicalResult verifySpec(::mlir::Location loc);

Verifies the validity of the specification and reports any errors at the given location. NOTE: This method must be implemented by the user.

Data Layout Scoping Operations ¶

Operations that define a scope for data layout queries, and that can be used to create a DataLayout object, are expected to implement the DataLayoutOpInterface. Such ops must provide at least a way of obtaining the data layout specification. The specification need not be necessarily attached to the operation as an attribute and may be constructed on-the-fly; it is only fetched once per DataLayout object and cached. Such ops may also provide custom handlers for data layout queries that provide results without forwarding the queries down to specific types or post-processing the results returned by types in target- or scope-specific ways. These custom handlers make it possible for scoping operations to (re)define data layout properties for types without having to modify the types themselves, e.g., when types are defined in another dialect.

DataLayoutOpInterface (`DataLayoutOpInterface`) ¶

Interface for operations that can have a data layout specification attached.

The DataLayout object, which can be used for data layout queries, can be constructed for such operations. The absence of a data layout specification must be handled without failing.

Concrete operations must implement the hook returning the data layout specification. They may optionally override the methods used in data layout queries, default implementations of which provide predefined answers for built-in types and dispatch to the type interface for all other types. These methods must be idempotent, that is return the same result on repeated queries with the same parameters. They are declared static and therefore have no access to the operation or its attributes. Instead, they receive a list of data layout entries relevant to the request. The entries are known to have passed the spec and entry verifier.

Methods: ¶

`getDataLayoutSpec` ¶

::mlir::DataLayoutSpecInterface getDataLayoutSpec();

Returns the data layout specification for this op, or null if it does not exist. NOTE: This method must be implemented by the user.

`getTypeSize` ¶

static ::llvm::TypeSize getTypeSize(::mlir::Type type, const ::mlir::DataLayout &dataLayout, ::mlir::DataLayoutEntryListRef params);

Returns the size of the given type computed using the relevant entries. The data layout object can be used for recursive queries. NOTE: This method must be implemented by the user.

`getTypeSizeInBits` ¶

static ::llvm::TypeSize getTypeSizeInBits(::mlir::Type type, const ::mlir::DataLayout &dataLayout, ::mlir::DataLayoutEntryListRef params);

Returns the size of the given type in bits computed using the relevant entries. The data layout object can be used for recursive queries. NOTE: This method must be implemented by the user.

`getTypeABIAlignment` ¶

static uint64_t getTypeABIAlignment(::mlir::Type type, const ::mlir::DataLayout &dataLayout, ::mlir::DataLayoutEntryListRef params);

Returns the alignment required by the ABI for the given type computed using the relevant entries. The data layout object can be used for recursive queries. NOTE: This method must be implemented by the user.

`getTypePreferredAlignment` ¶

static uint64_t getTypePreferredAlignment(::mlir::Type type, const ::mlir::DataLayout &dataLayout, ::mlir::DataLayoutEntryListRef params);

Returns the alignment preferred by the given type computed using the relevant entries. The data layoutobject can be used for recursive queries. NOTE: This method must be implemented by the user.

`getIndexBitwidth` ¶

static std::optional<uint64_t> getIndexBitwidth(::mlir::Type type, const ::mlir::DataLayout &dataLayout, ::mlir::DataLayoutEntryListRef params);

Returns the bitwidth that should be used when performing index computations for the type computed using the relevant entries. The data layout object can be used for recursive queries. NOTE: This method must be implemented by the user.

`getEndianness` ¶

static ::mlir::Attribute getEndianness(::mlir::DataLayoutEntryInterface entry);

Returns the endianness used by the ABI computed using the relevant entries. The data layout object can be used for recursive queries. NOTE: This method must be implemented by the user.

`getAllocaMemorySpace` ¶

static ::mlir::Attribute getAllocaMemorySpace(::mlir::DataLayoutEntryInterface entry);

Returns the memory space used by the ABI computed using the relevant entries. The data layout object can be used for recursive queries. NOTE: This method must be implemented by the user.

`getProgramMemorySpace` ¶

static ::mlir::Attribute getProgramMemorySpace(::mlir::DataLayoutEntryInterface entry);

Returns the memory space used by the ABI computed using the relevant entries. The data layout object can be used for recursive queries. NOTE: This method must be implemented by the user.

`getGlobalMemorySpace` ¶

static ::mlir::Attribute getGlobalMemorySpace(::mlir::DataLayoutEntryInterface entry);

Returns the memory space used by the ABI computed using the relevant entries. The data layout object can be used for recursive queries. NOTE: This method must be implemented by the user.

`getStackAlignment` ¶

static uint64_t getStackAlignment(::mlir::DataLayoutEntryInterface entry);

Returns the natural stack alignment in bits computed using the relevant entries. The data layout object can be used for recursive queries. NOTE: This method must be implemented by the user.

Types with Data Layout ¶

Type classes that intend to handle data layout queries themselves are expected to implement the DataLayoutTypeInterface. This interface provides overridable hooks for each data layout query. Each of these hooks is supplied with the type instance, a DataLayout object suitable for recursive queries, and a list of data layout queries relevant for the type class. It is expected to provide a valid result even if the list of entries is empty. These hooks do not have access to the operation in the scope of which the query is handled and should use the supplied entries instead.

DataLayoutTypeInterface (`DataLayoutTypeInterface`) ¶

Interface for types subject to data layout.

Types willing to be supported by the data layout subsystem should implement this interface by providing implementations of functions querying their size, required and preferred alignment. Each of these functions accepts as arguments a data layout object that can be used to perform recursive queries in the same scope, and a list of data layout entries relevant to this type. Specifically, the entries are those that have as key any instance of the same type class as the current type. For example, if IntegerType had implemented this interface, it would have received the entries with keys i1, i2, i8, etc. regardless of the bitwidth of this type. This mechanism allows types to “interpolate” the results in a type-specific way instead of listing all possible types in the specification.

The list of entries may be empty, in which case the type must provide a reasonable default value. The entries in the list are known to have passed the spec and the entry verifiers, as well as the type-specified verifier if provided.

In case of nested layout specs or spec changes, the type can override a hook indicating whether the outer (old) and the inner (new) spec are compatible.

Methods: ¶

`getTypeSize` ¶

::llvm::TypeSize getTypeSize(const ::mlir::DataLayout &dataLayout, ::mlir::DataLayoutEntryListRef params);

Returns the size of this type in bytes. NOTE: This method must be implemented by the user.

`getTypeSizeInBits` ¶

::llvm::TypeSize getTypeSizeInBits(const ::mlir::DataLayout &dataLayout, ::mlir::DataLayoutEntryListRef params);

Returns the size of this type in bits. NOTE: This method must be implemented by the user.

`getABIAlignment` ¶

uint64_t getABIAlignment(const ::mlir::DataLayout &dataLayout, ::mlir::DataLayoutEntryListRef params);

Returns the ABI-required alignment for this type, in bytes NOTE: This method must be implemented by the user.

`getPreferredAlignment` ¶

uint64_t getPreferredAlignment(const ::mlir::DataLayout &dataLayout, ::mlir::DataLayoutEntryListRef params);

Returns the preferred alignment for this type, in bytes. NOTE: This method must be implemented by the user.

`getIndexBitwidth` ¶

std::optional<uint64_t> getIndexBitwidth(const ::mlir::DataLayout &dataLayout, ::mlir::DataLayoutEntryListRef params);

Returns the bitwidth that should be used when performing index computations for the given pointer-like type. If the type is not a pointer-like type, returns std::nullopt. NOTE: This method must be implemented by the user.

`areCompatible` ¶

bool areCompatible(::mlir::DataLayoutEntryListRef oldLayout, ::mlir::DataLayoutEntryListRef newLayout);

Returns true if the two lists of entries are compatible, that is, that newLayout spec entries can be nested in an op with oldLayout spec entries. NOTE: This method must be implemented by the user.

`verifyEntries` ¶

::mlir::LogicalResult verifyEntries(::mlir::DataLayoutEntryListRef entries, ::mlir::Location loc);

Verifies that the given list of entries is valid for this type. NOTE: This method must be implemented by the user.

Dialects with Data Layout Identifiers ¶

For data layout entries that are not related to a particular type class, the key of the entry is an Identifier that belongs to some dialect. In this case, the dialect is expected to implement the DataLayoutDialectInterface. This dialect provides hooks for verifying the validity of the entry value attributes and for and the compatibility of nested entries.

Bits and Bytes ¶

Two versions of hooks are provided for sizes: in bits and in bytes. The version in bytes has a default implementation that derives the size in bytes by rounding up the result of division of the size in bits by 8. Types exclusively targeting architectures with different assumptions can override this. Operations can redefine this for all types, providing scoped versions for cases of byte sizes other than eight without having to modify types, including built-in types.

Query Dispatch ¶

The overall flow of a data layout property query is as follows.

The user constructs a DataLayout at the given scope. The constructor fetches the data layout specification and combines it with those of enclosing scopes (layouts are expected to be compatible).
The user calls DataLayout::query(Type ty).
If DataLayout has a cached response, this response is returned immediately.
Otherwise, the query is handed down by DataLayout to the closest layout scoping operation. If it implements DataLayoutOpInterface, then the query is forwarded toDataLayoutOpInterface::query(ty, *this, relevantEntries) where the relevant entries are computed as described above. If it does not implement DataLayoutOpInterface, it must be a ModuleOp, and the query is forwarded to DataLayoutTypeInterface::query(dataLayout, relevantEntries) after casting ty to the type interface.
Unless the query hook is reimplemented by the op interface, the query is handled further down to DataLayoutTypeInterface::query(dataLayout, relevantEntries) after casting ty to the type interface. If the type does not implement the interface, an unrecoverable fatal error is produced.
The type is expected to always provide the response, which is returned up the call stack and cached by the DataLayout.

Default Implementation ¶

The default implementation of the data layout interfaces directly handles queries for a subset of built-in types.

Built-in Modules ¶

Built-in ModuleOp allows at most one attribute that implements DataLayoutSpecInterface. It does not implement the entire interface for efficiency and layering reasons. Instead, DataLayout can be constructed for ModuleOp and handles modules transparently alongside other operations that implement the interface.

Built-in Types ¶

The following describes the default properties of built-in types.

The size of built-in integers and floats in bytes is computed as ceildiv(bitwidth, 8). The ABI alignment of integer types with bitwidth below 64 and of the float types is the closest from above power-of-two number of bytes. The ABI alignment of integer types with bitwidth 64 and above is 4 bytes (32 bits).

The size of built-in vectors is computed by first rounding their number of elements in the innermost dimension to the closest power-of-two from above, then getting the total number of elements, and finally multiplying it with the element size. For example, vector<3xi32> and vector<4xi32> have the same size. So do vector<2x3xf32> and vector<2x4xf32>, but vector<3x4xf32> and vector<4x4xf32> have different sizes. The ABI and preferred alignment of vector types is computed by taking the innermost dimension of the vector, rounding it up to the closest power-of-two, taking a product of that with element size in bytes, and rounding the result up again to the closest power-of-two.

Note: these values are selected for consistency with the default data layout in LLVM, which MLIR assumed until the introduction of proper data layout modeling, and with the modeling of n-D vectors. They may change in the future.

`index` type ¶

Index type is an integer type used for target-specific size information in, e.g., memref operations. Its data layout is parameterized by a single integer data layout entry that specifies its bitwidth. For example,

module attributes { dlti.dl_spec = #dlti.dl_spec<
  #dlti.dl_entry<index, 32>
>} {}

specifies that index has 32 bits and index computations should be performed using 32-bit precision as well. All other layout properties of index match those of the integer type with the same bitwidth defined above.

In absence of the corresponding entry, index is assumed to be a 64-bit integer.

`complex` type ¶

By default complex type is treated like a 2 element structure of its given element type. This is to say that each of its elements are aligned to their preferred alignment, the entire complex type is also aligned to this preference, and the complex type size includes the possible padding between elements to enforce alignment.

Byte Size ¶

The default data layout assumes 8-bit bytes.

DLTI Dialect ¶

The DLTI dialect provides the attributes implementing DataLayoutSpecInterface and DataLayoutEntryInterface, as well as a dialect attribute that can be used to attach the specification to a given operation. The verifier of this attribute triggers those of the specification and checks the compatibility of nested specifications.

Data Layout Modeling

Usage ¶

Scoping ¶

Compatibility and Transformations ¶

Queries ¶

Custom Implementations ¶

Data Layout Specifications ¶

DataLayoutEntryInterface (DataLayoutEntryInterface) ¶

Methods: ¶

getKey ¶

getValue ¶

verifyEntry ¶

DataLayoutSpecInterface (DataLayoutSpecInterface) ¶

Methods: ¶

combineWith ¶

getEntries ¶

getEndiannessIdentifier ¶

getAllocaMemorySpaceIdentifier ¶

getProgramMemorySpaceIdentifier ¶

getGlobalMemorySpaceIdentifier ¶

getStackAlignmentIdentifier ¶

getSpecForType ¶

getSpecForIdentifier ¶

verifySpec ¶

Data Layout Scoping Operations ¶

DataLayoutOpInterface (DataLayoutOpInterface) ¶

Methods: ¶

getDataLayoutSpec ¶

getTypeSize ¶

getTypeSizeInBits ¶

getTypeABIAlignment ¶

getTypePreferredAlignment ¶

getIndexBitwidth ¶

getEndianness ¶

getAllocaMemorySpace ¶

getProgramMemorySpace ¶

getGlobalMemorySpace ¶

getStackAlignment ¶

Types with Data Layout ¶

DataLayoutTypeInterface (DataLayoutTypeInterface) ¶

Methods: ¶

getTypeSize ¶

getTypeSizeInBits ¶

getABIAlignment ¶

getPreferredAlignment ¶

getIndexBitwidth ¶

areCompatible ¶

verifyEntries ¶

Dialects with Data Layout Identifiers ¶

Bits and Bytes ¶

Query Dispatch ¶

Default Implementation ¶

Built-in Modules ¶

Built-in Types ¶

index type ¶

complex type ¶

Byte Size ¶

DLTI Dialect ¶

DataLayoutEntryInterface (`DataLayoutEntryInterface`) ¶

`getKey` ¶

`getValue` ¶

`verifyEntry` ¶

DataLayoutSpecInterface (`DataLayoutSpecInterface`) ¶

`combineWith` ¶

`getEntries` ¶

`getEndiannessIdentifier` ¶

`getAllocaMemorySpaceIdentifier` ¶

`getProgramMemorySpaceIdentifier` ¶

`getGlobalMemorySpaceIdentifier` ¶

`getStackAlignmentIdentifier` ¶

`getSpecForType` ¶

`getSpecForIdentifier` ¶

`verifySpec` ¶

DataLayoutOpInterface (`DataLayoutOpInterface`) ¶

`getDataLayoutSpec` ¶

`getTypeSize` ¶

`getTypeSizeInBits` ¶

`getTypeABIAlignment` ¶

`getTypePreferredAlignment` ¶

`getIndexBitwidth` ¶

`getEndianness` ¶

`getAllocaMemorySpace` ¶

`getProgramMemorySpace` ¶

`getGlobalMemorySpace` ¶

`getStackAlignment` ¶

DataLayoutTypeInterface (`DataLayoutTypeInterface`) ¶

`getTypeSize` ¶

`getTypeSizeInBits` ¶

`getABIAlignment` ¶

`getPreferredAlignment` ¶

`getIndexBitwidth` ¶

`areCompatible` ¶

`verifyEntries` ¶

`index` type ¶

`complex` type ¶