[mlir][vector] Use getShapeForUnroll's default implementation.

Author: amd-eochoalo

mlir/include/mlir/Dialect/Vector/IR/VectorOps.td

@@ -2057,7 +2057,7 @@ def Vector_GatherOp :
   Vector_Op<"gather", [
     DeclareOpInterfaceMethods<MaskableOpInterface>,
     DeclareOpInterfaceMethods<MemorySpaceCastConsumerOpInterface>,
-    DeclareOpInterfaceMethods<VectorUnrollOpInterface, ["getShapeForUnroll"]>,
+    DeclareOpInterfaceMethods<VectorUnrollOpInterface>
     DeclareOpInterfaceMethods<AlignmentAttrOpInterface>
   ]>,
     Arguments<(ins Arg<TensorOrMemRef<[AnyType]>, "", [MemRead]>:$base,

mlir/lib/Dialect/Vector/IR/VectorOps.cpp

@@ -5982,10 +5982,6 @@ Type GatherOp::getExpectedMaskType() {
                          vecType.getScalableDims());
 }
 
-std::optional<SmallVector<int64_t, 4>> GatherOp::getShapeForUnroll() {
-  return llvm::to_vector<4>(getVectorType().getShape());
-}
-
 /// Cheeck if `indexVec` is constant 1D vec of consecutive values [0, 1, 2, ...]
 static LogicalResult isZeroBasedContiguousSeq(Value indexVec) {
   auto vecType = dyn_cast<VectorType>(indexVec.getType());