diff --git a/mlir/include/mlir/Dialect/MemRef/IR/MemRefOps.td b/mlir/include/mlir/Dialect/MemRef/IR/MemRefOps.td
--- a/mlir/include/mlir/Dialect/MemRef/IR/MemRefOps.td
+++ b/mlir/include/mlir/Dialect/MemRef/IR/MemRefOps.td
@@ -693,6 +693,68 @@
   let hasVerifier = 1;
 }
 
+//===----------------------------------------------------------------------===//
+// ExtractMetadataOp
+//===----------------------------------------------------------------------===//
+
+def MemRef_ExtractMetadataOp : MemRef_Op<"extract_metadata",
+    SameVariadicResultSize
+  ]> {
+  let summary = "Late binds a base buffer/offset/strides";
+  let description = [{
+    Queries a base buffer, offset and strides. This op is intended to be used 
+    late in the pipeline, allowing additional layers of transformations and 
+    foldings to be added as lowering progresses from higher-level dialect to
+    lower level dialects such as the LLVM dialect.
+
+    Its main use case is to lift the logic of ops that manipulate metadata from
+    the LLVM dialect and to make lowering more progressive. This brings the 
+    following benefits:
+      - foldings and canonicalizations can happen at a higher level in MLIR: 
+        before this op existed, lowering to LLVM would create large amounts of 
+        LLVMIR. Even when LLVM does a good job at folding the low-level IR from
+        a performance perspective, it is unnecessarily opaque and inefficient to
+        send unkempt IR to LLVM. This part could be solved by replicating LLVM 
+        foldings and canonicalizations in the LLVM dialect but this would still
+        not solve the next problems. 
+      - not all users of MLIR want to lower to LLVM and the information to e.g.
+        lower to library calls (e.g. libxsmm) or to SPIR-V is plainly not 
+        available.
+
+    This operation can be viewed as the natural complement to the
+    `memref.reinterpret_cast` op.
+
+    Example:
+
+    ```
+      %base, %offset, %sizes, %strides = extract_metadata %memref
+        : memref<10x?xf32>
+
+      // After folding, the type of %m2 can be memref<10x?xf32> and further 
+      // folded to %memref.
+      %m2 = memref.reinterpret_cast %base to
+          offset: [%offset],
+          sizes: [%sizes#0, %sizes#1],
+          strides: [%strides#0, %strides#1]
+        : memref<f32> to memref<?x?xf32, offset: ?, strides: [?, ?]>
+    ```
+  }];
+
+  let arguments = (ins
+    AnyStridedMemRef:$source
+  );
+  let results = (outs
+    AnyStridedMemRefOfRank<0>:$base_buffer,
+    Index:$offset,
+    Variadic<Index>:$sizes,
+    Variadic<Index>:$strides
+  );
+
+  let assemblyFormat = [{
+    $source `:` type($source) `->` type(results) attr-dict
+  }];
+}
+
 //===----------------------------------------------------------------------===//
 // GenericAtomicRMWOp
 //===----------------------------------------------------------------------===//