[Mlir-commits] [mlir] [MLIR][Linalg] Fix empty tensor assumptions for linalg.pack decomposition (PR #160246)

[Andrzej Warzyński]

================
@@ -57,7 +57,8 @@ class Linalg_RelayoutOp<string mnemonic, list<Trait> traits = []> :
     /// tile factors.
     DenseMap<int64_t, OpFoldResult> getDimAndTileMapping();
 
-    /// Return the tile sizes as OpFoldResult.
+    /// Return the tile sizes as OpFoldResult. Will return the Value
+    /// of the constant Op, not the constant Attribute.
----------------
banach-space wrote:

That in the following case, this method returns a `Value` corresponding to `%size` rather than a DenseAttribute with value `1`: 
```mlir
%size = arith.constant 1 : i32
```

It's probably worth clarifying further.

https://github.com/llvm/llvm-project/pull/160246