[mlir][Tensor] Update ParallelInsertSlicOp semantics to match that of InsertSliceOp

This revision updates the op semantics to also allow rank-reducing behavior as well
as updates the implementation

[mlir][Tensor] Update ParallelInsertSlicOp semantics to match that of InsertSliceOp

This revision updates the op semantics to also allow rank-reducing behavior as well
as updates the implementation to reuse code between the sequential and the parallel
version of the op.

Depends on D128920

Differential Revision: https://reviews.llvm.org/D128985

show more ...

[mlir][Tensor] Move ParallelInsertSlice to the tensor dialect

This is moslty NFC and will allow tensor.parallel_insert_slice to gain
rank-reducing semantics by reusing the vast majority of the tenso

[mlir][Tensor] Move ParallelInsertSlice to the tensor dialect

This is moslty NFC and will allow tensor.parallel_insert_slice to gain
rank-reducing semantics by reusing the vast majority of the tensor.insert_slice impl.

Depends on D128857

Differential Revision: https://reviews.llvm.org/D128920

show more ...

[mlir][Tensor][NFC] Better document rank-reducing behavior of ExtractSliceOp and cleanup

[mlir] Flip more uses to prefixed accessor form (NFC).

Try to keep the final flip small. Need to flip MemRef as there are many
templated cases with it and Tensor.

[mlir][sparse] accept sparse reshape (expand/collapse)

This revision makes sure we accept sparse tensors as arguments
of the expand/collapse reshaping operations in the tensor dialect.
Note that the

[mlir][sparse] accept sparse reshape (expand/collapse)

This revision makes sure we accept sparse tensors as arguments
of the expand/collapse reshaping operations in the tensor dialect.
Note that the actual lowering to runnable IR is still TBD.

Reviewed By: springerm

Differential Revision: https://reviews.llvm.org/D128311

show more ...

[mlir] Don't use Optional::getValue (NFC)

[mlir] Don't use Optional::hasValue (NFC)

[mlir] Start migrating more dialects to prefixed form

Marked all dialects that could be (reasonably) easily flipped to _Both
prefix. Updating the accessors to prefixed form will happen in follow
up,

[mlir] Start migrating more dialects to prefixed form

Marked all dialects that could be (reasonably) easily flipped to _Both
prefix. Updating the accessors to prefixed form will happen in follow
up, this was to flush out conflicts and to mark all dialects explicitly
as I plan to flip OpBase default to _Prefixed to avoid needing to
migrate new dialects.

Except for Standalone example which got flipped to _Prefixed.

Differential Revision: https://reviews.llvm.org/D128027

show more ...

[mlir][tensor] Add canonicalization for tensor.cast from extract_slice

Propagate static size information into extract_slice producer if
possible.

Differential Revision: https://reviews.llvm.org/D12

[mlir][tensor] Add canonicalization for tensor.cast from extract_slice

Propagate static size information into extract_slice producer if
possible.

Differential Revision: https://reviews.llvm.org/D125972

show more ...

[DenseElementAttr] Simplify the public API for creating these.

Instead of requiring the client to compute the "isSplat" bit,
compute it internally. This makes the logic more consistent
and defines

[DenseElementAttr] Simplify the public API for creating these.

Instead of requiring the client to compute the "isSplat" bit,
compute it internally. This makes the logic more consistent
and defines away a lot of "elements.size()==1" in the clients.

This addresses Issue #55185

Differential Revision: https://reviews.llvm.org/D125447

show more ...

[mlir][NFC] Shift a bunch of dialect includes from the .h to the .cpp

Now that dialect constructors are generated in the .cpp file, we can
drop all of the dependent dialect includes from the .h file

[mlir][NFC] Shift a bunch of dialect includes from the .h to the .cpp

Now that dialect constructors are generated in the .cpp file, we can
drop all of the dependent dialect includes from the .h file.

Differential Revision: https://reviews.llvm.org/D124298

show more ...

[mlir][tensor] Promote extract(from_elements(...)) to folding pattern

Differential Revision: https://reviews.llvm.org/D123617

[mlir][tensor] Add pattern to fold ExtractSliceOp, PadOp chains.

The pattern folds chains of tensor::ExtractSliceOp, tensor::PadOp pairs if they pad different dimensions. Repeated tiling and padding

[mlir][tensor] Add pattern to fold ExtractSliceOp, PadOp chains.

The pattern folds chains of tensor::ExtractSliceOp, tensor::PadOp pairs if they pad different dimensions. Repeated tiling and padding of the tiled dimensions may introduce such chains. This canonicalization pattern folds these chains to a single tensor::ExtractSliceOp, tensor::PadOp pair that pads all dimensions at once, which simplifies vectorization and bufferization.

Example:
```mlir
%0 = tensor.extract_slice %input[16, 0] [%sz0, 64] [1, 1]
: tensor<64x64xf32> to tensor<?x64xf32>
%1 = tensor.pad %0 low[0, 0] high[%pw0, 0] { ...
} : tensor<?x64xf32> to tensor<8x64xf32>
%2 = tensor.extract_slice %1[0, 4] [8, %sz1] [1, 1]
: tensor<8x64xf32> to tensor<8x?xf32>
%res = tensor.pad %2 nofold low[0, 0] high[0, %pw1] { ...
} : tensor<8x?xf32> to tensor<8x4xf32>
```
folds into:
```mlir
%0 = tensor.extract_slice %input[16, 4] [%sz0, %sz1] [1, 1]
: tensor<64x64xf32> to tensor<?x?xf32>
%res = tensor.pad %0 nofold low[0, 0] high[%pw0, %pw1] { ...
} : tensor<?x?xf32> to tensor<8x4xf32>
```

Reviewed By: nicolasvasilache, hanchung

Differential Revision: https://reviews.llvm.org/D122722

show more ...

Revert "Revert "[mlir] Rewrite canonicalization of collapse(expand) and expand(collapse).""

This reverts commit 96e9b6c9dc60946f08399def879a19395bc98107.

Revert "[mlir] Rewrite canonicalization of collapse(expand) and expand(collapse)."

This reverts commit 64f659bee67b5a024defeb3cd2ecf65e1ad8c0a7.

An invalid tensor.expand_shape op is generated with

Revert "[mlir] Rewrite canonicalization of collapse(expand) and expand(collapse)."

This reverts commit 64f659bee67b5a024defeb3cd2ecf65e1ad8c0a7.

An invalid tensor.expand_shape op is generated with the commit. To repro:

$ mlir-opt -canonicalize a.mlir

```
func @foo(%0: tensor<1x1xf32>, %1: tensor<1x1xf32>, %2: tensor<1x1xf32>) -> tensor<1x1xf32> {
%cst = arith.constant 0.000000e+00 : f32
%3 = linalg.init_tensor [8, 1] : tensor<8x1xf32>
%4 = linalg.fill ins(%cst : f32) outs(%3 : tensor<8x1xf32>) -> tensor<8x1xf32>
%5 = tensor.collapse_shape %0 [] : tensor<1x1xf32> into tensor<f32>
%6 = tensor.insert_slice %5 into %4[0, 0] [1, 1] [1, 1] : tensor<f32> into tensor<8x1xf32>
%7 = linalg.init_tensor [8, 1] : tensor<8x1xf32>
%8 = linalg.fill ins(%cst : f32) outs(%7 : tensor<8x1xf32>) -> tensor<8x1xf32>
%9 = tensor.collapse_shape %2 [] : tensor<1x1xf32> into tensor<f32>
%10 = tensor.insert_slice %9 into %8[0, 0] [1, 1] [1, 1] : tensor<f32> into tensor<8x1xf32>
%11 = tensor.collapse_shape %6 [[0, 1]] : tensor<8x1xf32> into tensor<8xf32>
%12 = linalg.init_tensor [8] : tensor<8xf32>
%13 = linalg.generic {indexing_maps = [affine_map<(d0) -> (d0)>, affine_map<(d0) -> (d0)>], iterator_types = ["parallel"]} ins(%11 : tensor<8xf32>) outs(%12 : tensor<8xf32>) {
^bb0(%arg3: f32, %arg4: f32):
linalg.yield %arg3 : f32
} -> tensor<8xf32>
%14 = tensor.expand_shape %13 [[0, 1, 2, 3]] : tensor<8xf32> into tensor<1x1x8x1xf32>
%15 = tensor.collapse_shape %1 [] : tensor<1x1xf32> into tensor<f32>
%16 = linalg.init_tensor [] : tensor<f32>
%17 = linalg.generic {indexing_maps = [affine_map<() -> ()>, affine_map<() -> ()>], iterator_types = []} ins(%15 : tensor<f32>) outs(%16 : tensor<f32>) {
^bb0(%arg3: f32, %arg4: f32):
linalg.yield %arg3 : f32
} -> tensor<f32>
%18 = tensor.expand_shape %17 [] : tensor<f32> into tensor<1x1x1x1xf32>
%19 = tensor.collapse_shape %10 [[0, 1]] : tensor<8x1xf32> into tensor<8xf32>
%20 = linalg.init_tensor [8] : tensor<8xf32>
%21 = linalg.generic {indexing_maps = [affine_map<(d0) -> (d0)>, affine_map<(d0) -> (d0)>], iterator_types = ["parallel"]} ins(%19 : tensor<8xf32>) outs(%20 : tensor<8xf32>) {
^bb0(%arg3: f32, %arg4: f32):
linalg.yield %arg3 : f32
} -> tensor<8xf32>
%22 = tensor.expand_shape %21 [[0, 1, 2, 3]] : tensor<8xf32> into tensor<1x1x8x1xf32>
%23 = linalg.mmt4d {comment = "f32*f32->f32, aarch64, matrix*vector"} ins(%14, %18 : tensor<1x1x8x1xf32>, tensor<1x1x1x1xf32>) outs(%22 : tensor<1x1x8x1xf32>) -> tensor<1x1x8x1xf32>
%24 = tensor.collapse_shape %23 [[0, 1, 2, 3]] : tensor<1x1x8x1xf32> into tensor<8xf32>
%25 = linalg.init_tensor [8] : tensor<8xf32>
%26 = linalg.generic {indexing_maps = [affine_map<(d0) -> (d0)>, affine_map<(d0) -> (d0)>], iterator_types = ["parallel"]} ins(%24 : tensor<8xf32>) outs(%25 : tensor<8xf32>) {
^bb0(%arg3: f32, %arg4: f32):
linalg.yield %arg3 : f32
} -> tensor<8xf32>
%27 = tensor.expand_shape %26 [[0, 1]] : tensor<8xf32> into tensor<8x1xf32>
%28 = tensor.extract_slice %27[0, 0] [1, 1] [1, 1] : tensor<8x1xf32> to tensor<f32>
%29 = tensor.expand_shape %28 [] : tensor<f32> into tensor<1x1xf32>
return %29 : tensor<1x1xf32>
}
```

Differential Revision: https://reviews.llvm.org/D123161

show more ...

[mlir] Rewrite canonicalization of collapse(expand) and expand(collapse).

Differential Revision: https://reviews.llvm.org/D122666

[mlir] Rename `OpAsmParser::OperandType` to `OpAsmParser::UnresolvedOperand`

I am not sure about the meaning of Type in the name (was it meant be interpreted as Kind?), and given the importance and

[mlir] Rename `OpAsmParser::OperandType` to `OpAsmParser::UnresolvedOperand`

I am not sure about the meaning of Type in the name (was it meant be interpreted as Kind?), and given the importance and meaning of Type in the context of MLIR, its probably better to rename it. Given the comment in the source code, the suggestion in the GitHub issue and the final discussions in the review, this patch renames the OperandType to UnresolvedOperand.

Fixes https://github.com/llvm/llvm-project/issues/54446

Differential Revision: https://reviews.llvm.org/D122142

show more ...

[mlir] Remove incorrect builders for ExpandShapeOp

ExpandShapeOp builder cannot infer the result type since it doesn't know
how the dimension needs to be split. Remove this builder so that it
doesn'

[mlir] Remove incorrect builders for ExpandShapeOp

ExpandShapeOp builder cannot infer the result type since it doesn't know
how the dimension needs to be split. Remove this builder so that it
doesn't get used accidently. Also remove one potential path using it in
generic fusion.

Differential Revision: https://reviews.llvm.org/D122019

show more ...

[mlir][tensor] Implement ReifyRankedShapedTypeOpInterface on GenerateOp

Differential Revision: https://reviews.llvm.org/D121520

[mlir] Support verification order (3/3)

In this CL, update the function name of verifier according to the
behavior. If a verifier needs to access the region then it'll be updated
to `verifyRegions`.

[mlir] Support verification order (3/3)

In this CL, update the function name of verifier according to the
behavior. If a verifier needs to access the region then it'll be updated
to `verifyRegions`.

Reviewed By: rriddle

Differential Revision: https://reviews.llvm.org/D120373

show more ...

[mlir] Add canonicalizations for op -> tensor.cast folding.

A `tensor.cast` consumer can be folded with its producer. This is
beneficial only if the result of the tensor cast is more static than
the

[mlir] Add canonicalizations for op -> tensor.cast folding.

A `tensor.cast` consumer can be folded with its producer. This is
beneficial only if the result of the tensor cast is more static than
the source. This patch adds a utility function to check that this is
the case, and adds a couple of canonicalizations patterns that fold an
operation with `tensor.cast` conusmers.

Reviewed By: gysit

Differential Revision: https://reviews.llvm.org/D120950

show more ...

[mlir] Match Arithmetic::ConstantOp and Tensor::ExtractSliceOp.

Add a pattern matcher for ExtractSliceOp when its source is a constant.

The matching heuristics can be governed by the control functi

[mlir] Match Arithmetic::ConstantOp and Tensor::ExtractSliceOp.

Add a pattern matcher for ExtractSliceOp when its source is a constant.

The matching heuristics can be governed by the control function since
generating a new constant is not always beneficial.

Differential Revision: https://reviews.llvm.org/D119605

show more ...

Revert "[mlir] Fold Arithmetic::ConstantOp and Tensor::ExtractSliceOp."

This reverts commit 3104994104f0c2f274acf5e01eb6cc82e9cca06b.

[mlir] Fold Arithmetic::ConstantOp and Tensor::ExtractSliceOp.

Fold ExtractSliceOp when the source is a constant.

Fold Tensor.extract_slice into a constant splat.

Fold arith.extract_slice into arith.constant when the source is a constant
splat and the result type is statically shaped.