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

[lhutton1]

This commit allows reshape to be created with an unranked output, allowing it to be inferred by the shape inference pass.

[llvmbot]

@llvm/pr-subscribers-mlir-tosa

@llvm/pr-subscribers-mlir

Author: Luke Hutton (lhutton1)

Changes

This commit allows reshape to be created with an unranked output, allowing it to be inferred by the shape inference pass.

---

Full diff: https://github.com/llvm/llvm-project/pull/140617.diff

3 Files Affected:

• (modified) mlir/include/mlir/Dialect/Tosa/IR/TosaOps.td (+1-1)
• (modified) mlir/lib/Dialect/Tosa/IR/TosaOps.cpp (+8-5)
• (modified) mlir/test/Dialect/Tosa/ops.mlir (+8)

diff --git a/mlir/include/mlir/Dialect/Tosa/IR/TosaOps.td b/mlir/include/mlir/Dialect/Tosa/IR/TosaOps.td
index 52bb0eb992b69..86f9ab94ec152 100644
--- a/mlir/include/mlir/Dialect/Tosa/IR/TosaOps.td
+++ b/mlir/include/mlir/Dialect/Tosa/IR/TosaOps.td
@@ -1959,7 +1959,7 @@ def Tosa_ReshapeOp : Tosa_InferTensorTypeOp<"reshape"> {
   );
 
   let results = (outs
-    Tosa_RankedTensor:$output
+    Tosa_Tensor:$output
   );
 
   list<Availability> availability = [
diff --git a/mlir/lib/Dialect/Tosa/IR/TosaOps.cpp b/mlir/lib/Dialect/Tosa/IR/TosaOps.cpp
index b2e471f2bba93..b74b820e11f75 100644
--- a/mlir/lib/Dialect/Tosa/IR/TosaOps.cpp
+++ b/mlir/lib/Dialect/Tosa/IR/TosaOps.cpp
@@ -2012,7 +2012,6 @@ llvm::LogicalResult tosa::ReshapeOp::verify() {
     return failure();
   }
   TensorType inputType = getInput1().getType();
-  RankedTensorType outputType = getType();
 
   SmallVector<int64_t> shapeValues;
   if (!tosa::getConstShapeValues(getShape().getDefiningOp(), shapeValues)) {
@@ -2020,6 +2019,14 @@ llvm::LogicalResult tosa::ReshapeOp::verify() {
     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";
 
@@ -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();
 }
 
diff --git a/mlir/test/Dialect/Tosa/ops.mlir b/mlir/test/Dialect/Tosa/ops.mlir
index f8273190bde40..e727614bd76f9 100644
--- a/mlir/test/Dialect/Tosa/ops.mlir
+++ b/mlir/test/Dialect/Tosa/ops.mlir
@@ -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> {