[Mlir-commits] [mlir] [MLIR][XeGPU] Use the `setupDpasLayout` utility for dpas layout propagation (PR #180937)

[Jianhui Li]

================
@@ -852,4 +852,209 @@ xegpu::setupStoreMatrixAnchorLayout(xegpu::LayoutKind layoutKind,
 
   return setupGenericStoreAnchorLayout(layoutKind, context, false, maxChunkSize,
                                        srcShape, subgroupSize);
-}
\ No newline at end of file
+}
+
+// This function returns the default lane layout for a given vector type.
+// - `packingSize` means multiple consecutive elements can be accessed together
+// as a single unit.
+// - `vnni` means data packing is column-wise (i.e., 2x1xf16 with vnni vs.
+// 1x2xf16 w/o vnni).
+template <typename RankedTy>
+static xegpu::LayoutAttr
+getDefaultLaneLayout(RankedTy ty, const xegpu::uArch::uArch *uArch,
----------------
Jianhui-Li wrote:

It doesn't apply to all ops (like xegpu.layout doesn't necessary follow this).  
Can you add qualifier to the name - like getDefault2DBlockLaneLayout ? 

https://github.com/llvm/llvm-project/pull/180937