[Mlir-commits] [mlir] [mlir][xegpu] Remove chunk_size attribute; infer from types (PR #205122)

[Jianhui Li]

================
@@ -1871,6 +1867,9 @@ static xegpu::DistributeLayoutAttr setupGenericLoadAnchorLayout(
   SmallVector<int64_t> consumerLaneData =
       consumerLayout.getEffectiveLaneDataAsInt();
 
+  // Take lane_layout / lane_data from the consumer. (See the TODO above: the
+  // `maxChunkSize`-capped derive path used by the store side is not yet needed
+  // here, but will be for matrix coalescing.)
----------------
Jianhui-Li wrote:

//capped derive path used by the store side is not yet needed here, but will be for matrix coalescing

not sure what the comments mean?  Be careful if it is AI-generated 

https://github.com/llvm/llvm-project/pull/205122