diff --git a/mlir/include/mlir/Dialect/Linalg/IR/LinalgNamedStructuredOps.yaml b/mlir/include/mlir/Dialect/Linalg/IR/LinalgNamedStructuredOps.yaml
--- a/mlir/include/mlir/Dialect/Linalg/IR/LinalgNamedStructuredOps.yaml
+++ b/mlir/include/mlir/Dialect/Linalg/IR/LinalgNamedStructuredOps.yaml
@@ -9,6 +9,8 @@
 
     Numeric casting is performed on the input operand, promoting it to the same
     data type as the accumulator/output.
+  defines:
+  - hasCanonicalizer
 structured_op: !LinalgStructuredOpConfig
   args:
   - !LinalgOperandDefConfig
diff --git a/mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp b/mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp
--- a/mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp
+++ b/mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp
@@ -224,11 +224,11 @@
 
   if (addOperandSegmentSizes) {
     // This is a bit complex because we're trying to be backward compatible with
-    // operation syntax that mix the inherent attributes and the discardable ones
-    // in the same dictionary.
-    // If the properties are used, we append the operand_segment_sizes there directly.
-    // Otherwise we append it to the discardable attributes dictionary where it is
-    // handled by the generic Operation::create(...) method.
+    // operation syntax that mix the inherent attributes and the discardable
+    // ones in the same dictionary. If the properties are used, we append the
+    // operand_segment_sizes there directly. Otherwise we append it to the
+    // discardable attributes dictionary where it is handled by the generic
+    // Operation::create(...) method.
     if (result.propertiesAttr) {
       NamedAttrList attrs = llvm::cast<DictionaryAttr>(result.propertiesAttr);
       attrs.append("operand_segment_sizes",
@@ -539,6 +539,33 @@
 
 } // namespace
 
+//===----------------------------------------------------------------------===//
+// CopyOp
+//===----------------------------------------------------------------------===//
+
+namespace {
+
+struct EraseSelfCopyOnBuffers : OpRewritePattern<CopyOp> {
+  using OpRewritePattern<CopyOp>::OpRewritePattern;
+  LogicalResult matchAndRewrite(CopyOp copyOp,
+                                PatternRewriter &rewriter) const override {
+    if (!copyOp.hasBufferSemantics())
+      return rewriter.notifyMatchFailure(copyOp,
+                                         "does not have buffer semantics");
+    if (copyOp.getInputs().front() != copyOp.getOutputs().front())
+      return rewriter.notifyMatchFailure(copyOp, "not a self copy");
+    rewriter.eraseOp(copyOp);
+    return success();
+  }
+};
+
+} // namespace
+
+void CopyOp::getCanonicalizationPatterns(RewritePatternSet &results,
+                                         MLIRContext *context) {
+  results.add<EraseSelfCopyOnBuffers>(context);
+}
+
 //===----------------------------------------------------------------------===//
 // FillOp
 //===----------------------------------------------------------------------===//
@@ -2114,8 +2141,7 @@
   for (unsigned i = 0; i < sourceShape.size(); i++) {
     int64_t dimShape = sourceShape[i];
     AffineExpr dimExpr = sourceMap.getResult(i);
-    if (!affineExprToSize.contains(dimExpr) ||
-        !sourceType.isDynamicDim(i)) {
+    if (!affineExprToSize.contains(dimExpr) || !sourceType.isDynamicDim(i)) {
       newShape.push_back(dimShape);
       continue;
     }
diff --git a/mlir/python/mlir/dialects/linalg/opdsl/ops/core_named_ops.py b/mlir/python/mlir/dialects/linalg/opdsl/ops/core_named_ops.py
--- a/mlir/python/mlir/dialects/linalg/opdsl/ops/core_named_ops.py
+++ b/mlir/python/mlir/dialects/linalg/opdsl/ops/core_named_ops.py
@@ -17,6 +17,7 @@
     Numeric casting is performed on the input operand, promoting it to the same
     data type as the accumulator/output.
     """
+    defines(Canonicalizer)
     O[None] = cast(U, I[None])
 
 
diff --git a/mlir/test/Dialect/Linalg/canonicalize.mlir b/mlir/test/Dialect/Linalg/canonicalize.mlir
--- a/mlir/test/Dialect/Linalg/canonicalize.mlir
+++ b/mlir/test/Dialect/Linalg/canonicalize.mlir
@@ -47,6 +47,16 @@
 
 // -----
 
+func.func @dce_self_linalg_copy(%arg0 : memref<?xf32>) {
+  linalg.copy ins(%arg0: memref<?xf32>) outs(%arg0: memref<?xf32>)
+  return
+}
+
+// CHECK-LABEL: @dce_self_linalg_copy
+//   CHECK-NOT:   copy
+
+// -----
+
 // CHECK-LABEL: func @tensor.cast(
 func.func @tensor.cast(%a : tensor<3x4xf32>, %b : tensor<4x?xf32>, %c : tensor<3x?xf32>)
   -> tensor<3x?xf32>