[llvm] [mlir][bufferization] Empty tensor elimination based on SubsetOpInterface (PR #65766)

[Martin Erhart via llvm-commits]

================
@@ -109,19 +109,17 @@ def EliminateEmptyTensorsOp
          DeclareOpInterfaceMethods<MemoryEffectsOpInterface>]> {
   let description = [{
     Try to eliminate all `tensor.empty` ops within the targeted op by replacing
-    them with a destination tensor.
+    them with another destination tensor.
 
-    `tensor.empty` ops cannot be bufferizes. They can either be converted to
-    `bufferization.alloc_tensor` or replaced with another tensor (via this
-    transform). `tensor.empty` does not specify the contents of the returned
+    "tensor.empty" ops cannot be bufferized. They can either be converted to
+    "bufferization.alloc_tensor" or replaced with another tensor (via this
+    transform). "tensor.empty" does not specify the contents of the returned
----------------
maerhart wrote:

Interesting, I often do the exact opposite. Is there a written guideline somewhere?

https://github.com/llvm/llvm-project/pull/65766