Please use GitHub pull requests for new patches. Phabricator shutdown timeline

Differential D133911 Diff 460287 mlir/include/mlir/Dialect/MemRef/IR/MemRefOps.td

Changeset View

Standalone View

mlir/include/mlir/Dialect/MemRef/IR/MemRefOps.td

Show First 20 Lines • Show All 315 Lines • ▼ Show 20 Lines	let description = [{
```		```
operation ::= ssa-id `=` `memref.cast` ssa-use `:` type `to` type		operation ::= ssa-id `=` `memref.cast` ssa-use `:` type `to` type
```		```

The `memref.cast` operation converts a memref from one type to an equivalent		The `memref.cast` operation converts a memref from one type to an equivalent
type with a compatible shape. The source and destination types are		type with a compatible shape. The source and destination types are
compatible if:		compatible if:

a. Both are ranked memref types with the same element type, address space,		a. Both memrefs have compatible element types: element type is the same or
and rank and:		differs only it the signed/signless/unsigned bit.

		b. Both are ranked memref types with compatible element types, address
		space, and rank and:
1. Both have the same layout or both have compatible strided layouts.		1. Both have the same layout or both have compatible strided layouts.
2. The individual sizes (resp. offset and strides in the case of strided		2. The individual sizes (resp. offset and strides in the case of strided
memrefs) may convert constant dimensions to dynamic dimensions and		memrefs) may convert constant dimensions to dynamic dimensions and
vice-versa.		vice-versa.

If the cast converts any dimensions from an unknown to a known size, then it		If the cast converts any dimensions from an unknown to a known size, then it
acts as an assertion that fails at runtime if the dynamic dimensions		acts as an assertion that fails at runtime if the dynamic dimensions
disagree with resultant destination size.		disagree with resultant destination size.
Show All 12 Lines	let description = [{
%4 = memref.cast %1 : memref<12x4xf32, strided<[?, ?], offset: ?>> to		%4 = memref.cast %1 : memref<12x4xf32, strided<[?, ?], offset: ?>> to
memref<12x4xf32, strided<[4, 1], offset: 5>>		memref<12x4xf32, strided<[4, 1], offset: 5>>
// Erase static offset and stride information, replacing it with		// Erase static offset and stride information, replacing it with
// dynamic information.		// dynamic information.
%5 = memref.cast %1 : memref<12x4xf32, strided<[4, 1], offset: 5>> to		%5 = memref.cast %1 : memref<12x4xf32, strided<[4, 1], offset: 5>> to
memref<12x4xf32, strided<[?, ?], offset: ?>>		memref<12x4xf32, strided<[?, ?], offset: ?>>
```		```

b. Either or both memref types are unranked with the same element type, and		c. Either or both memref types are unranked with compatible element type,
address space.		and the same address space.

Example:		Example:

```mlir		```mlir
Cast to concrete shape.		Cast to concrete shape.
%4 = memref.cast %1 : memref<*xf32> to memref<4x?xf32>		%4 = memref.cast %1 : memref<*xf32> to memref<4x?xf32>

Erase rank information.		Erase rank information.
▲ Show 20 Lines • Show All 341 Lines • ▼ Show 20 Lines	let description = [{
of transformations and foldings to be added as lowering progresses from		of transformations and foldings to be added as lowering progresses from
higher-level dialect to lower-level dialects such as the LLVM dialect.		higher-level dialect to lower-level dialects such as the LLVM dialect.

The op requires a strided memref source operand. If the source operand is not		The op requires a strided memref source operand. If the source operand is not
a strided memref, then verification fails.		a strided memref, then verification fails.

This operation is also useful for completeness to the existing memref.dim op.		This operation is also useful for completeness to the existing memref.dim op.
While accessing strides, offsets and the base pointer independently is not		While accessing strides, offsets and the base pointer independently is not
available, this is useful for composing with its natural complement op:		available, this is useful for composing with its natural complement op:
`memref.reinterpret_cast`.		`memref.reinterpret_cast`.

Intended Use Cases:		Intended Use Cases:

The main use case is to expose the logic for manipulate memref metadata at a		The main use case is to expose the logic for manipulate memref metadata at a
higher level than the LLVM dialect.		higher level than the LLVM dialect.
This makes lowering more progressive and brings the following benefits:		This makes lowering more progressive and brings the following benefits:
- not all users of MLIR want to lower to LLVM and the information to e.g.		- not all users of MLIR want to lower to LLVM and the information to e.g.
lower to library calls---like libxsmm---or to SPIR-V was not available.		lower to library calls---like libxsmm---or to SPIR-V was not available.
- foldings and canonicalizations can happen at a higher level in MLIR:		- foldings and canonicalizations can happen at a higher level in MLIR:
before this op existed, lowering to LLVM would create large amounts of		before this op existed, lowering to LLVM would create large amounts of
LLVMIR. Even when LLVM does a good job at folding the low-level IR from		LLVMIR. Even when LLVM does a good job at folding the low-level IR from
a performance perspective, it is unnecessarily opaque and inefficient to		a performance perspective, it is unnecessarily opaque and inefficient to
send unkempt IR to LLVM.		send unkempt IR to LLVM.

Example:		Example:

```mlir		```mlir
%base, %offset, %sizes:2, %strides:2 =		%base, %offset, %sizes:2, %strides:2 =
memref.extract_strided_metadata %memref :		memref.extract_strided_metadata %memref :
memref<10x?xf32>, index, index, index, index, index		memref<10x?xf32>, index, index, index, index, index

// After folding, the type of %m2 can be memref<10x?xf32> and further		// After folding, the type of %m2 can be memref<10x?xf32> and further
// folded to %memref.		// folded to %memref.
%m2 = memref.reinterpret_cast %base to		%m2 = memref.reinterpret_cast %base to
offset: [%offset],		offset: [%offset],
sizes: [%sizes#0, %sizes#1],		sizes: [%sizes#0, %sizes#1],
strides: [%strides#0, %strides#1]		strides: [%strides#0, %strides#1]
: memref<f32> to memref<?x?xf32, offset: ?, strides: [?, ?]>		: memref<f32> to memref<?x?xf32, offset: ?, strides: [?, ?]>
```		```
}];		}];
▲ Show 20 Lines • Show All 1,201 Lines • Show Last 20 Lines