[Mlir-commits] [mlir] [mlir][memref] Introduce `memref.distinct_objects` op (PR #156913)

[Matthias Springer]

================
@@ -168,14 +168,49 @@ def AssumeAlignmentOp : MemRef_Op<"assume_alignment", [
   let assemblyFormat = "$memref `,` $alignment attr-dict `:` type($memref)";
   let extraClassDeclaration = [{
     MemRefType getType() { return ::llvm::cast<MemRefType>(getResult().getType()); }
-    
+
     Value getViewSource() { return getMemref(); }
   }];
 
   let hasVerifier = 1;
   let hasFolder = 1;
 }
 
+//===----------------------------------------------------------------------===//
+// DistinctObjectsOp
+//===----------------------------------------------------------------------===//
+
+def DistinctObjectsOp : MemRef_Op<"distinct_objects", [
+      Pure,
+      DeclareOpInterfaceMethods<InferTypeOpInterface>
+      // ViewLikeOpInterface TODO: ViewLikeOpInterface only supports a single argument
+    ]> {
+  let summary = "assumption that acesses to specific memrefs will never alias";
+  let description = [{
+      The `distinct_objects` operation takes a list of memrefs and returns a list of
+      memrefs of the same types, with the additional assumption that accesses to
+      these memrefs will never alias with each other. This means that loads and
+      stores to different memrefs in the list can be safely reordered.
+
+      If the memrefs do alias, the behavior is undefined. This operation doesn't
+      affect the semantics of a program where the non-aliasing assumption holds
----------------
matthias-springer wrote:

I find this sentence confusing. Can this be rephrased as: `This operation doesn't affect the semantics of the program.` ?


https://github.com/llvm/llvm-project/pull/156913