[mlir] Allow Tile transform op to take dynamic sizes

[ftynse]

Summary

Extend the definition of the Tile structured transform op to enable it
accepting handles to operations that produce tile sizes at runtime. This is
useful by itself and prepares for more advanced tiling strategies. Note that
the changes are relevant only to the transform dialect, the tiling
transformation itself already supports dynamic sizes.

Depends On D129216

[nicolasvasilache]

Could you please rebase on https://reviews.llvm.org/D129185 ?

[ftynse]

Fix include.

[ftynse]

I rebased on the requested commit, but this is a non-functional change. If
rebasing implied using applyToOne, it is not applicable because TileOp now
takes several operand handles, which is not supported by applyToOne.

[ftynse]

Rebase.

[nicolasvasilache]

Nice improvements!

mlir/include/mlir/Dialect/Linalg/TransformOps/LinalgTransformOps.td

415

Can you please add a pedantic `

Return modes:
=============

doc section that resembles the other ones and describes all handles manipulation to a 5-y-o ?

423

seems like you want a SmallVector<OpFoldResult> getMixedSizes or something here to uniformize usage?

mlir/lib/Dialect/Linalg/TransformOps/LinalgTransformOps.cpp

185

does this need an error message or is this reasonably handled by errors from below ?

633

The fact that 0 encodes "do not tile along this dim" would ideally remain deeper and not leak up to here.
If we really want to manipulate at this level, we could use an interface but given you're just resizing here, I am wondering whether you can just rewrite:

for (const auto &en2 : llvm::enumerate(tiledOp->loops))
  loops[en2.index()].push_back(en2.value());

to something that only uses tiledOp->loops and which has already taken care of 0.

If that does not work it means that we have significant trickiness in the impl and much more doc is required (and prob an interface is warranted too)

635

SmallVector<Operation*> and friends usually resolve to 6, do you really care about setting a 4 here or does it fail to compile for some reason ?

653

this would moslty go away if you had a SmallVector<OpFoldResult>getMixedSizes and you would just
do vector of materializeOpFoldResult.

699

Ok I see you are actually making use of that 0 magic number knowledge.
In that case we should find a way to add it to the TilingInterface and use that more pervasively.

[nicolasvasilache]

In D129217#3635796, @ftynse wrote:

I rebased on the requested commit, but this is a non-functional change. If
rebasing implied using applyToOne, it is not applicable because TileOp now
takes several operand handles, which is not supported by applyToOne.

Nope just rebase because that change was somewhat tricky to land and potentially had merge conflicts

[ftynse]

Address review.

mlir/lib/Dialect/Linalg/TransformOps/LinalgTransformOps.cpp

185

It probably needs a better message, but this is out of the scope of this patch: the code is in FuseOp and is not touched by this commit.

635

Just decreasing the stack usage a bit here, the default-size computation was not designed for that.

653

It's more involved than that. The Value returned by getMixedSizes is the handle, not the payload op that produces the size, it cannot be just materialized.

699

Except that this tiling implementation doesn't use the interface.

[nicolasvasilache]

mlir/lib/Dialect/Linalg/TransformOps/LinalgTransformOps.cpp

699

doesn't *yet* use the interface :)

mlir/include/mlir/Dialect/Linalg/TransformOps/LinalgTransformOps.td

::mlir::DiagnosedSilenceableFailure applyToOne(
         ::mlir::linalg::LinalgOp target, 
         ::llvm::SmallVectorImpl<::mlir::Operation *> &results, 
         ::mlir::transform::TransformState &state);
   }];
 }
 
 def TileOp : Op<Transform_Dialect, "structured.tile",
        [DeclareOpInterfaceMethods<TransformOpInterface>,
-        FunctionalStyleTransformOpTrait, MemoryEffectsOpInterface]> {
+        DeclareOpInterfaceMethods<MemoryEffectsOpInterface>]> {
   let description = [{
-    Indicates that the given `target` op should be tiled with the options
-    provided as attributes. This transform generates a loop nest with a smaller
-    ("tiled") target operation in its body. Currently limited to LinalgOps.
-
-    `sizes` are the tile sizes. A tile size of `0` indicates that the
-    respective dimension should not be tiled. No loop will be generated for such
-    dimensions. If all tile sizes are `0`, this transform is effectively a
-    no-op.
+    Indicates that the given `target` op should be tiled with the given sizes.
+    This transform generates a loop nest with a smaller ("tiled") target
+    operation in its body. Currently limited to LinalgOps.
+
+    Tile sizes may be known at transformation time, in which case they are
+    expected to be provided in the `static_size` attribute, or not, in which
+    case the tile value must be computed by the payload IR and the handle to the
+    operation computing it must be provided through `dynamic_sizes`. When the
+    sizes are not known statically, the corresponding entry in the
+    `static_sizes` attribute must be set to `ShapedType::kDynamicSize`. Only
+    the dynamic sizes must be provided in `dynamic_sizes`, i.e., there should
+    be as many handles as `ShapedType::kDynamicSize` values in the
+    `static_sizes` attribute. A static size of `0` indicates that the dimension
+    should not be tiled. No loop will be generated for such dimensions. If all
+    tile sizes are `0`, this transform is effectively a no-op.
 
     This op returns handles to the tiled op (in the generated loop nest) and the
     generated loops. The number of loops is the number of non-zero tile sizes.
   }];
 
   let arguments = (ins PDL_Operation:$target,
-                   DefaultValuedAttr<I64ArrayAttr, "{}">:$sizes,
+                   Variadic<PDL_Operation>:$dynamic_sizes,
+                   DefaultValuedAttr<I64ArrayAttr, "{}">:$static_sizes,
                    DefaultValuedAttr<I64ArrayAttr, "{}">:$interchange);
   let results = (outs PDL_Operation:$tiled_linalg_op,
                       Variadic<PDL_Operation>:$loops);
 
   let hasCustomAssemblyFormat = 1;
 }
 
 def VectorizeOp : Op<Transform_Dialect, "structured.vectorize",

mlir/include/mlir/Dialect/Linalg/Utils/Utils.h

 /// they were originally extracted from with `extract_slice` before being passed
 /// as `operands` to the given structured operation `op` or its clone. Note that
 /// `operands` are not necessarily the actual operands of `op`, the operation
 /// serves only as metadata container for operand types and positions.
 SmallVector<Value> insertSlicesBack(OpBuilder &builder, Location loc,
                                     LinalgOp op, ValueRange operands,
                                     ValueRange results);
 
+/// Turns an OpFoldResult into a value, creating an index-typed constant if
+/// necessary.
+Value materializeOpFoldResult(ImplicitLocOpBuilder &builder,
+                              OpFoldResult opFoldResult);
+
 /// Creates an extract_slice/subview op for a single `valueToTile` with
 /// `builder`. This new operation extracts a tile of `valueToTile`, starting
 /// at offsets `lbs` and with sizes `subShapeSizes`. `omitPartialTileCheck`
 /// controls whether to omit the partial/boundary tile condition check in cases
 /// where we statically know that it is unnecessary.
 Value makeTiledShape(OpBuilder &builder, Location loc, Value valueToTile,
                      ValueRange tileSizes, AffineMap map, ValueRange lbs,
                      ValueRange ubs, ValueRange subShapeSizes,

mlir/lib/Dialect/Linalg/TransformOps/LinalgTransformOps.cpp

 //===- LinalgTransformOps.cpp - Implementation of Linalg transform ops ----===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/Linalg/TransformOps/LinalgTransformOps.h"
 
+#include "mlir/Dialect/Arithmetic/IR/Arithmetic.h"
 #include "mlir/Dialect/Linalg/IR/Linalg.h"
 #include "mlir/Dialect/Linalg/Transforms/Transforms.h"
 #include "mlir/Dialect/PDL/IR/PDL.h"
 #include "mlir/Dialect/PDL/IR/PDLTypes.h"
 #include "mlir/Dialect/Transform/IR/TransformDialect.h"
 #include "mlir/Parser/Parser.h"
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
 

 
 //===----------------------------------------------------------------------===//
 // FuseOp
 //===----------------------------------------------------------------------===//
 
 /// Apply a tiling transformation to all payload ops and store both the
 /// tiled operation as well as the created tile loops.
 static LogicalResult
-applyTilingToAll(Operation *transformOp, Value target,
-                 ArrayRef<int64_t> tileSizes,
+applyTilingToAll(Operation *transformOp, ArrayRef<Operation *> payloadOps,
+                 unsigned numLoops,
                  transform::TransformResults &transformResults,
-                 transform::TransformState &state,
                  function_ref<FailureOr<TiledLinalgOp>(LinalgOp)> applyFn) {
-  // Number of loops: Number of tiles sizes that are not zero.
-  size_t numLoops = tileSizes.size() - llvm::count(tileSizes, 0);
-  // All payload ops. These should all be LinalgOps for now.
-  ArrayRef<Operation *> payloadOps = state.getPayloadOps(target);
-
   SmallVector<Operation *> tiledLinalgOps;
   SmallVector<SmallVector<Operation *>> loopOps(numLoops);
   for (unsigned int i = 0; i < numLoops; ++i)
     loopOps[i].reserve(payloadOps.size());
 
   for (Operation *target : payloadOps) {
     auto linalgOp = dyn_cast<linalg::LinalgOp>(target);
     if (!linalgOp)

 DiagnosedSilenceableFailure
 transform::FuseOp::apply(mlir::transform::TransformResults &transformResults,
                          mlir::transform::TransformState &state) {
   LinalgTilingAndFusionOptions fusionOptions;
   fusionOptions.tileSizes = extractI64Array(getTileSizes());
   fusionOptions.tileInterchange = extractI64Array(getTileInterchange());
 
   LogicalResult result = applyTilingToAll(
-      getOperation(), getTarget(), fusionOptions.tileSizes, transformResults,
-      state, [&](LinalgOp linalgOp) -> FailureOr<TiledLinalgOp> {
+      getOperation(), state.getPayloadOps(getTarget()),
+      fusionOptions.tileSizes.size() - llvm::count(fusionOptions.tileSizes, 0),
+      transformResults, [&](LinalgOp linalgOp) -> FailureOr<TiledLinalgOp> {
         LinalgTileAndFuseTensorOpsPattern pattern(getContext(), fusionOptions);
         SimpleRewriter rewriter(getContext());
         rewriter.setInsertionPoint(linalgOp);
         FailureOr<TileLoopNest> tileLoopNest =
             pattern.returningMatchAndRewrite(linalgOp, rewriter);
         if (failed(tileLoopNest))
           return failure();
 
         TiledLinalgOp tiledLinalgOp;
         tiledLinalgOp.op = tileLoopNest->getRootOp();
         tiledLinalgOp.loops = {tileLoopNest->getLoopOps().begin(),
                                tileLoopNest->getLoopOps().end()};
         return tiledLinalgOp;
       });
-  return failed(result) ? DiagnosedSilenceableFailure::definiteFailure()
-                        : DiagnosedSilenceableFailure::success();
+  return DiagnosedSilenceableFailure(result);
 }
 
 ParseResult transform::FuseOp::parse(OpAsmParser &parser,
                                      OperationState &result) {
   return parseTileLikeOp(
       parser, result,
       transform::FuseOp::getTileSizesAttrName(result.name).getValue());
 }

 //===----------------------------------------------------------------------===//
 // TileOp
 //===----------------------------------------------------------------------===//
 
 DiagnosedSilenceableFailure
 transform::TileOp::apply(TransformResults &transformResults,
                          TransformState &state) {
   LinalgTilingOptions tilingOptions;
-  SmallVector<int64_t> tileSizes = extractI64Array(getSizes());
+  SmallVector<int64_t> tileSizes = extractI64Array(getStaticSizes());
+
+  ArrayRef<Operation *> targets = state.getPayloadOps(getTarget());
+  SmallVector<ArrayRef<Operation *>> dynamicSizeProducers;
+  dynamicSizeProducers.reserve(getDynamicSizes().size());
+  for (Value dynamicSizeProducerHandle : getDynamicSizes()) {
+    dynamicSizeProducers.push_back(
+        state.getPayloadOps(dynamicSizeProducerHandle));
+
+    if (dynamicSizeProducers.back().size() != targets.size()) {
+      DiagnosedSilenceableFailure diag =
+          emitSilenceableError()
+          << "expected as many dynamic size-producing operations ("
+          << dynamicSizeProducers.back().size() << ") as target ops ("
+          << targets.size() << ")";
+      diag.attachNote(dynamicSizeProducerHandle.getLoc()) << "for this handle";
+      return diag;
+    }
+
+    for (Operation *op : dynamicSizeProducers.back()) {
+      if (op->getNumResults() == 1 &&
+          op->getResult(0).getType().isa<IndexType>())
+        continue;
+      DiagnosedSilenceableFailure diag =
+          emitSilenceableError() << "expected sizes to be produced by ops "
+                                    "with a single index-type result";
+      diag.attachNote(op->getLoc()) << "size producer op";
+      diag.attachNote(dynamicSizeProducerHandle.getLoc()) << "for this handle";
+      return diag;
+    }
+  }
+
+  unsigned numLoops = tileSizes.size() - llvm::count(tileSizes, 0);
+  SmallVector<Operation *> tiled;
+  SmallVector<SmallVector<Operation *, 4>, 4> loops;
+  loops.resize(numLoops);
+  for (auto &en : llvm::enumerate(targets)) {
+    auto linalgOp = dyn_cast<LinalgOp>(en.value());
+    if (!linalgOp) {
+      DiagnosedSilenceableFailure diag = emitSilenceableError()
+                                         << "only linalg ops are supported";
+      diag.attachNote(en.value()->getLoc()) << "target op";
+      return diag;
+    }
+
+    unsigned index = en.index();
+    if (!tileSizes.empty()) {
+      tilingOptions.setTileSizeComputationFunction([&, index](OpBuilder &b,
+                                                              Operation *) {
+        SmallVector<Value, 4> sizes;
+        sizes.reserve(tileSizes.size());
+        unsigned dynamicIdx = 0;
+        for (int64_t size : tileSizes) {
+          if (size == ShapedType::kDynamicSize) {
+            sizes.push_back(
+                dynamicSizeProducers[dynamicIdx++][index]->getResult(0));
+          } else {
+            sizes.push_back(b.create<arith::ConstantIndexOp>(getLoc(), size));
+          }
+        }
+        return sizes;
+      });
+    }
 
-  if (!tileSizes.empty())
-    tilingOptions.setTileSizes(tileSizes);
     tilingOptions.setInterchange(extractUIntArray(getInterchange()));
     LinalgTilingPattern pattern(getContext(), tilingOptions);
-
-  LogicalResult result = applyTilingToAll(
-      getOperation(), getTarget(), tileSizes, transformResults, state,
-      [&](LinalgOp linalgOp) {
-        SimpleRewriter rewriter(linalgOp.getContext());
-        return pattern.returningMatchAndRewrite(linalgOp, rewriter);
-      });
-  return DiagnosedSilenceableFailure(result);
+    FailureOr<TiledLinalgOp> tiledOp =
+        pattern.returningMatchAndRewrite(linalgOp, rewriter);
+    if (failed(tiledOp))
+      return DiagnosedSilenceableFailure::definiteFailure();
+
+    tiled.push_back(tiledOp->op);
+    for (const auto &en2 : llvm::enumerate(tiledOp->loops))
+      loops[en2.index()].push_back(en2.value());
+  }
+
+  transformResults.set(getTiledLinalgOp().cast<OpResult>(), tiled);
+  for (const auto &en : llvm::enumerate(loops))
+    transformResults.set(getLoops()[en.index()].cast<OpResult>(), en.value());
+
+  return DiagnosedSilenceableFailure::success();
 }
 
 ParseResult transform::TileOp::parse(OpAsmParser &parser,
                                      OperationState &result) {
-  return parseTileLikeOp(parser, result,
-                         TileOp::getSizesAttrName(result.name).getValue());
+  OpAsmParser::UnresolvedOperand target;
+  SmallVector<OpAsmParser::UnresolvedOperand> dynamicSizes;
+  ArrayAttr staticSizes;
+  auto pdlOperationType = pdl::OperationType::get(parser.getContext());
+  if (parser.parseOperand(target) ||
+      parser.resolveOperand(target, pdlOperationType, result.operands) ||
+      parseOperandsOrIntegersSizesList(parser, dynamicSizes, staticSizes) ||
+      parser.resolveOperands(dynamicSizes, pdlOperationType, result.operands) ||
+      parser.parseOptionalAttrDict(result.attributes))
+    return ParseResult::failure();
+
+  result.addAttribute(getStaticSizesAttrName(result.name), staticSizes);
+  size_t numExpectedLoops =
+      staticSizes.size() - llvm::count(extractI64Array(staticSizes), 0);
+  result.addTypes(SmallVector<Type>(numExpectedLoops + 1, pdlOperationType));
+  return success();
 }
 
 void TileOp::print(OpAsmPrinter &p) {
-  p << ' ';
-  p << getTarget();
-  p.printOptionalAttrDict((*this)->getAttrs());
+  p << ' ' << getTarget();
+  printOperandsOrIntegersSizesList(p, getOperation(), getDynamicSizes(),
+                                   getStaticSizes());
+  p.printOptionalAttrDict((*this)->getAttrs(), {getStaticSizesAttrName()});
+}
+
+void transform::TileOp::getEffects(
+    SmallVectorImpl<MemoryEffects::EffectInstance> &effects) {
+  consumesHandle(getTarget(), effects);
+  onlyReadsHandle(getDynamicSizes(), effects);
+  producesHandle(getTiledLinalgOp(), effects);
+  producesHandle(getLoops(), effects);
+  modifiesPayload(effects);
 }
 
 //===----------------------------------------------------------------------===//
 // VectorizeOp
 //===----------------------------------------------------------------------===//
 
 DiagnosedSilenceableFailure
 transform::VectorizeOp::applyToOne(Operation *target,

 namespace {
 /// Registers new ops and declares PDL as dependent dialect since the additional
 /// ops are using PDL types for operands and results.
 class LinalgTransformDialectExtension
     : public transform::TransformDialectExtension<
           LinalgTransformDialectExtension> {
 public:
   LinalgTransformDialectExtension() {
+    declareDependentDialect<arith::ArithmeticDialect>();
     declareDependentDialect<pdl::PDLDialect>();
     declareDependentDialect<scf::SCFDialect>();
     declareDependentDialect<vector::VectorDialect>();
     registerTransformOps<
 #define GET_OP_LIST
 #include "mlir/Dialect/Linalg/TransformOps/LinalgTransformOps.cpp.inc"
         >();
   }

mlir/lib/Dialect/Linalg/Transforms/Split.cpp

 #include "mlir/Dialect/Linalg/Transforms/Transforms.h"
 #include "mlir/Dialect/Linalg/Utils/Utils.h"
 
 #include "llvm/ADT/STLExtras.h"
 
 using namespace mlir;
 using namespace mlir::linalg;
 
-/// Turns an OpFoldResult into a value, creating an index-typed constant if
-/// necessary.
-static Value materializeOpFoldResult(ImplicitLocOpBuilder &builder,
-                                     OpFoldResult opFoldResult) {
-  if (opFoldResult.is<Value>())
-    return opFoldResult.get<Value>();
-  auto attr = opFoldResult.get<Attribute>().cast<IntegerAttr>();
-  return builder.create<arith::ConstantIndexOp>(attr.getValue().getSExtValue());
-}
-
 /// Extract the slices of `operands` supplied to the given operation `op` such
 /// that they are sufficient to execute the op for the subset of its iteration
 /// space defined by `splitIterationSpace`. The subset is a part of the original
 /// iteration space split at the given `dimension`. If `offset` is provided, it
 /// indicates the iterator value at which the dimension has been split and
 /// requires the "high" part starting at the given offset of the operands to be
 /// generated; otherwise, the "low" part with no offset is generated. Note that
 /// `operands` are not necessarily the actual operands of `op`.

mlir/lib/Dialect/Linalg/Utils/Utils.cpp

for (OpOperand *opOperand : op.getOutputTensorOperands()) {
     } else {
       tensorResults.push_back(results[resultIdx]);
     }
     ++resultIdx;
   }
   return tensorResults;
 }
 
+Value materializeOpFoldResult(ImplicitLocOpBuilder &builder,
+                              OpFoldResult opFoldResult) {
+  if (auto value = opFoldResult.dyn_cast<Value>())
+    return value;
+  auto attr = opFoldResult.get<Attribute>().cast<IntegerAttr>();
+  return builder.create<arith::ConstantIndexOp>(attr.getValue().getSExtValue());
+}
+
 SmallVector<Value, 4> makeTiledShapes(OpBuilder &b, Location loc,
                                       LinalgOp linalgOp,
                                       ArrayRef<Value> valuesToTile,
                                       ValueRange ivs, ValueRange tileSizes,
                                       ArrayRef<Value> sizeBounds,
                                       bool omitPartialTileCheck) {
   assert(ivs.size() == static_cast<size_t>(llvm::count_if(
                            llvm::make_range(tileSizes.begin(), tileSizes.end()),

mlir/python/mlir/dialects/_structured_transform_ops_ext.py

 
 
 class TileOp:
   """Specialization for TileOp class."""
 
   def __init__(self,
                target: Union[Operation, Value],
                *,
-               sizes: OptionalIntList = None,
+               sizes: Optional[Union[Sequence[Union[int, IntegerAttr, Operation,
+                                                    Value]], ArrayAttr]] = None,
                interchange: OptionalIntList = None,
                loc=None,
                ip=None):
     pdl_operation_type = pdl.OperationType.get()
-    sizes_attr = _get_int_array_attr(sizes)
+    i64_type = IntegerType.get_signless(64)
+
+    if sizes is None:
+      sizes = []
+
+    static_sizes = []
+    dynamic_sizes = []
+    if isinstance(sizes, ArrayAttr):
+      sizes_attr = sizes
+    else:
+      for size in sizes:
+        if isinstance(size, int):
+          static_sizes.append(IntegerAttr.get(i64_type, size))
+        elif isinstance(size, IntegerAttr):
+          static_sizes.append(size)
+        else:
+          static_sizes.append(
+              IntegerAttr.get(i64_type, ShapedType._get_dynamic_size()))
+          dynamic_sizes.append(_get_op_result_or_value(size))
+      sizes_attr = ArrayAttr.get(static_sizes)
+
     num_loops = sum(
         v if v == 0 else 1 for v in self.__extract_values(sizes_attr))
     super().__init__(
         pdl_operation_type, [pdl_operation_type] * num_loops,
         _get_op_result_or_value(target),
-        sizes=sizes_attr,
+        dynamic_sizes=dynamic_sizes,
+        static_sizes=sizes_attr,
         interchange=_get_int_array_attr(interchange) if interchange else None,
         loc=loc,
         ip=ip)
 
   def __extract_values(self, attr: Optional[ArrayAttr]) -> List[int]:
     if not attr:
       return []
     return [IntegerAttr(element).value for element in attr]

mlir/test/Dialect/Linalg/transform-op-scalarize.mlir

pdl.pattern @pdl_target : benefit(1) {
     %0 = pdl.operation "linalg.matmul"(%args : !pdl.range<value>) -> (%results : !pdl.range<type>)
     // TODO: we don't want this, but it is the required terminator for pdl.pattern
     rewrite %0 with "transform.dialect"
   }
 
   transform.sequence %arg0 {
   ^bb1(%arg1: !pdl.operation):
     %0 = pdl_match @pdl_target in %arg1
-    %1, %loops = transform.structured.tile %0 {sizes = [10, 0, 0]}
+    %1, %loops = transform.structured.tile %0 [10, 0, 0]
     %2 = transform.structured.scalarize %1
   }
 }

mlir/test/Dialect/Linalg/transform-op-tile.mlir

-// RUN: mlir-opt --test-transform-dialect-interpreter %s | FileCheck %s
+// RUN: mlir-opt --test-transform-dialect-interpreter --split-input-file %s | FileCheck %s
 
 transform.with_pdl_patterns {
 ^bb0(%arg0: !pdl.operation):
   sequence %arg0 {
     ^bb0(%arg1: !pdl.operation):
       %0 = pdl_match @pdl_target in %arg1
-      %1, %loops:3 = transform.structured.tile %0 {sizes = [4, 4, 4]}
+      %1, %loops:3 = transform.structured.tile %0 [4, 4, 4]
   }
 
   pdl.pattern @pdl_target : benefit(1) {
     %args = operands
     %results = types
     %0 = operation "linalg.matmul"(%args : !pdl.range<value>) -> (%results : !pdl.range<type>)
     rewrite %0 with "transform.dialect"
   }
//      CHECK:   scf.yield %[[TD1]] : tensor<128x128xf32>
   %0 = linalg.matmul  ins(%arg0, %arg1: tensor<128x128xf32>, tensor<128x128xf32>)
                      outs(%arg2: tensor<128x128xf32>)
     -> tensor<128x128xf32>
 
 //      CHECK: return %[[TD0]] : tensor<128x128xf32>
   return %0 : tensor<128x128xf32>
 }
 
+// -----
+
+transform.with_pdl_patterns {
+^bb0(%arg0: !pdl.operation):
+  sequence %arg0 {
+    ^bb0(%arg1: !pdl.operation):
+      %0 = pdl_match @pdl_target in %arg1
+      %1 = pdl_match @func_call in %arg1
+      %2, %loops:3 = transform.structured.tile %0 [%1, %1, 4]
+  }
+
+  pdl.pattern @pdl_target : benefit(1) {
+    %args = operands
+    %results = types
+    %0 = operation "linalg.matmul"(%args : !pdl.range<value>) -> (%results : !pdl.range<type>)
+    rewrite %0 with "transform.dialect"
+  }
+  pdl.pattern @func_call : benefit(1) {
+    %args = operands
+    %results = types
+    %0 = operation "func.call"(%args : !pdl.range<value>) -> (%results : !pdl.range<type>)
+    rewrite %0 with "transform.dialect"
+  }
+}
+
+func.func private @get_dynamic_tile_size() -> index
+
+// CHECK-LABEL: func @tile_linalg_matmul_dynamic(
+// CHECK-SAME:    %[[TA:[0-9a-z]+]]: tensor<128x128xf32>
+// CHECK-SAME:    %[[TB:[0-9a-z]+]]: tensor<128x128xf32>
+// CHECK-SAME:    %[[TC:[0-9a-z]+]]: tensor<128x128xf32>
+// CHECK-SAME:  -> tensor<128x128xf32> {
+func.func @tile_linalg_matmul_dynamic(
+  %arg0: tensor<128x128xf32>, %arg1: tensor<128x128xf32>, %arg2: tensor<128x128xf32>)
+    -> tensor<128x128xf32> {
+//      CHECK: %[[TD0:.*]] = scf.for {{.*}} to {{.*}} step {{.*}} iter_args(%[[TC0:.*]] = %[[TC]]) -> (tensor<128x128xf32>) {
+//      CHECK:   %[[TD1:.*]] = scf.for {{.*}} to {{.*}} step {{.*}} iter_args(%[[TC1:.*]] = %[[TC0]]) -> (tensor<128x128xf32>) {
+//      CHECK:     %[[TD2:.*]] = scf.for {{.*}} to {{.*}} step {{.*}} iter_args(%[[TC2:.*]] = %[[TC1]]) -> (tensor<128x128xf32>) {
+//      CHECK:       %[[sTA:.*]] = tensor.extract_slice %[[TA]][{{.*}}] : tensor<128x128xf32> to tensor<?x4xf32>
+//      CHECK:       %[[sTB:.*]] = tensor.extract_slice %[[TB]][{{.*}}] : tensor<128x128xf32> to tensor<4x?xf32>
+//      CHECK:       %[[sTC:.*]] = tensor.extract_slice %[[TC2]][{{.*}}] : tensor<128x128xf32> to tensor<?x?xf32>
+//      CHECK:       %[[sTD:.*]] = linalg.matmul ins(%[[sTA]], %[[sTB]] : tensor<?x4xf32>, tensor<4x?xf32>)
+// CHECK-SAME:                                   outs(%[[sTC]] : tensor<?x?xf32>)  -> tensor<?x?xf32>
+//      CHECK:       %[[TD:.*]] = tensor.insert_slice %[[sTD]] into %[[TC2]][{{.*}}]  : tensor<?x?xf32> into tensor<128x128xf32>
+//      CHECK:       scf.yield %[[TD]] : tensor<128x128xf32>
+//      CHECK:     scf.yield %[[TD2]] : tensor<128x128xf32>
+//      CHECK:   scf.yield %[[TD1]] : tensor<128x128xf32>
+  %sz = func.call @get_dynamic_tile_size() : () -> index
+  %0 = linalg.matmul  ins(%arg0, %arg1: tensor<128x128xf32>, tensor<128x128xf32>)
+                     outs(%arg2: tensor<128x128xf32>)
+    -> tensor<128x128xf32>
+
+//      CHECK: return %[[TD0]] : tensor<128x128xf32>
+  return %0 : tensor<128x128xf32>
+}

mlir/test/Dialect/Linalg/transform-ops.mlir

 // RUN: mlir-opt %s | mlir-opt | FileCheck %s
 
 transform.sequence {
 ^bb1(%arg0: !pdl.operation):
   // CHECK %{{.*}}, %{{.*}}:2 = transform.structured.tile
-  %0, %1:2 = transform.structured.tile %arg0 { sizes = [2, 0, 3] }
+  %0, %1:2 = transform.structured.tile %arg0 [2, 0, 3]
 }
 
 //===----------------------------------------------------------------------===//
 // Check that operations are registered correctly through the extension
 // mechanism. Their syntax is generated and requries no additional testing since
 // we test the generator.
 //===----------------------------------------------------------------------===//
 

mlir/test/Dialect/Transform/selective-targeting.mlir

pdl.pattern @pdl_target_attrC : benefit(1) {
     %0 = operation "linalg.matmul"(%args : !pdl.range<value>) {"test.attrC" = %attr}-> (%results : !pdl.range<type>)
     // TODO: we don't want this, but it is the required terminator for pdl.pattern
     rewrite %0 with "transform.dialect"
   }
 
   transform.sequence %arg0 {
   ^bb1(%arg1: !pdl.operation):
     %0 = pdl_match @pdl_target_attrA in %arg1
-    transform.structured.tile %0 {sizes = [4, 4, 4]}
+    transform.structured.tile %0 [4, 4, 4]
     %1 = pdl_match @pdl_target_attrC in %arg1
     %2 = transform.get_closest_isolated_parent %1
     transform.structured.vectorize %2
   }
 }
 
 // -----
 

mlir/test/python/dialects/transform_structured_ext.py

 @run
 def testTileCompact():
   sequence = transform.SequenceOp()
   with InsertionPoint(sequence.body):
     structured.TileOp(sequence.bodyTarget, sizes=[4, 8], interchange=[0, 1])
     transform.YieldOp()
   # CHECK-LABEL: TEST: testTileCompact
   # CHECK: transform.sequence
-  # CHECK: %{{.+}}, %{{.+}}:2 = transform.structured.tile
-  # CHECK-DAG: interchange = [0, 1]
-  # CHECK-DAG: sizes = [4, 8]
+  # CHECK: %{{.+}}, %{{.+}}:2 = transform.structured.tile %{{.*}}[4, 8]
+  # CHECK: interchange = [0, 1]
 
 
 @run
 def testTileAttributes():
   sequence = transform.SequenceOp()
   attr = ArrayAttr.get(
       [IntegerAttr.get(IntegerType.get_signless(64), x) for x in [4, 8]])
   ichange = ArrayAttr.get(
       [IntegerAttr.get(IntegerType.get_signless(64), x) for x in [0, 1]])
   with InsertionPoint(sequence.body):
     structured.TileOp(sequence.bodyTarget, sizes=attr, interchange=ichange)
     transform.YieldOp()
   # CHECK-LABEL: TEST: testTileAttributes
   # CHECK: transform.sequence
-  # CHECK: structured.tile
-  # CHECK-DAG: interchange = [0, 1]
-  # CHECK-DAG: sizes = [4, 8]
+  # CHECK: %{{.+}}, %{{.+}}:2 = transform.structured.tile %{{.*}}[4, 8]
+  # CHECK: interchange = [0, 1]
 
 
 @run
 def testTileZero():
   sequence = transform.SequenceOp()
   with InsertionPoint(sequence.body):
     structured.TileOp(
         sequence.bodyTarget, sizes=[4, 0, 2, 0], interchange=[0, 1, 2, 3])
     transform.YieldOp()
   # CHECK-LABEL: TEST: testTileZero
   # CHECK: transform.sequence
-  # CHECK: %{{.+}}, %{{.+}}:2 = transform.structured.tile
-  # CHECK-DAG: interchange = [0, 1, 2, 3]
-  # CHECK-DAG: sizes = [4, 0, 2, 0]
+  # CHECK: %{{.+}}, %{{.+}}:2 = transform.structured.tile %{{.*}}[4, 0, 2, 0]
+  # CHECK: interchange = [0, 1, 2, 3]
+
+
+@run
+def testTileDynamic():
+  with_pdl = transform.WithPDLPatternsOp()
+  with InsertionPoint(with_pdl.body):
+    sequence = transform.SequenceOp(with_pdl.bodyTarget)
+    with InsertionPoint(sequence.body):
+      m1 = transform.PDLMatchOp(sequence.bodyTarget, "first")
+      m2 = transform.PDLMatchOp(sequence.bodyTarget, "second")
+      structured.TileOp(sequence.bodyTarget, sizes=[m1, 3, m2, 0])
+      transform.YieldOp()
+  # CHECK-LABEL: TEST: testTileDynamic
+  # CHECK: %[[FIRST:.+]] = pdl_match
+  # CHECK: %[[SECOND:.+]] = pdl_match
+  # CHECK: %{{.+}}, %{{.+}}:3 = transform.structured.tile %{{.*}}[%[[FIRST]], 3, %[[SECOND]], 0]
 
 
 @run
 def testVectorize():
   sequence = transform.SequenceOp()
   with InsertionPoint(sequence.body):
     structured.VectorizeOp(sequence.bodyTarget, vectorize_padding=True)
     transform.YieldOp()
   # CHECK-LABEL: TEST: testVectorize
   # CHECK: transform.sequence
   # CHECK: = transform.structured.vectorize
   # CHECK: vectorize_padding = true

utils/bazel/llvm-project-overlay/mlir/BUILD.bazel

cc_library(
     srcs = [
         "lib/Dialect/Linalg/TransformOps/LinalgTransformOps.cpp",
     ],
     hdrs = [
         "include/mlir/Dialect/Linalg/TransformOps/LinalgTransformOps.h",
     ],
     includes = ["include"],
     deps = [
+        ":ArithmeticDialect",
         ":IR",
         ":LinalgDialect",
         ":LinalgTransformOpsIncGen",
         ":LinalgTransforms",
         ":PDLDialect",
         ":Parser",
         ":SideEffectInterfaces",
         ":TransformDialect",