Skip to content

[mlir][tosa] Allow creation of reshape with unranked output #140617

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
May 21, 2025
Merged

[mlir][tosa] Allow creation of reshape with unranked output #140617

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion mlir/include/mlir/Dialect/Tosa/IR/TosaOps.td
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1959,7 +1959,7 @@ def Tosa_ReshapeOp : Tosa_InferTensorTypeOp<"reshape"> {
);

let results = (outs
Tosa_RankedTensor:$output
Tosa_Tensor:$output
);

list<Availability> availability = [
Expand Down
13 changes: 8 additions & 5 deletions mlir/lib/Dialect/Tosa/IR/TosaOps.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2012,14 +2012,21 @@ llvm::LogicalResult tosa::ReshapeOp::verify() {
return failure();
}
TensorType inputType = getInput1().getType();
RankedTensorType outputType = getType();

SmallVector<int64_t> shapeValues;
if (!tosa::getConstShapeValues(getShape().getDefiningOp(), shapeValues)) {
// skip following checks if shape is not constant
return mlir::success();
}

int missingDims = llvm::count(shapeValues, -1);
if (missingDims > 1)
return emitOpError() << "expected at most one target dimension to be -1";

const auto outputType = dyn_cast<RankedTensorType>(getType());
if (!outputType)
return success();

if ((int64_t)shapeValues.size() != outputType.getRank())
return emitOpError() << "new shape does not match result rank";

Expand Down Expand Up @@ -2056,10 +2063,6 @@ llvm::LogicalResult tosa::ReshapeOp::verify() {
}
}

int missingDims = llvm::count(shapeValues, -1);
if (missingDims > 1)
return emitOpError() << "expected at most one target dimension to be -1";

return mlir::success();
}

Expand Down
8 changes: 8 additions & 0 deletions mlir/test/Dialect/Tosa/ops.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -643,6 +643,14 @@ func.func @test_reshape(%arg0: tensor<13x21x3xf32>) -> tensor<1x819xf32> {
return %0 : tensor<1x819xf32>
}

// -----
// CHECK-LABEL: reshape_unranked_output
func.func @test_reshape_unranked_output(%arg0: tensor<13x21x3xf32>) -> tensor<*xf32> {
%1 = tosa.const_shape {values = dense<[21, 13, 3]> : tensor<3xindex>} : () -> !tosa.shape<3>
%0 = tosa.reshape %arg0, %1 : (tensor<13x21x3xf32>, !tosa.shape<3>) -> tensor<*xf32>
return %0 : tensor<*xf32>
}

// -----
// CHECK-LABEL: reverse
func.func @test_reverse(%arg0: tensor<13x21x3xf32>) -> tensor<13x21x3xf32> {
Expand Down