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

[Jianhui Li]

================
@@ -758,8 +753,9 @@ def XeGPU_LoadGatherOp : XeGPU_Op<"load", [MemoryEffects<[MemRead]>, AnchorLayou
   ```
 
   Example 2 (lane level):
-  lane level only accepts the offsets variant. chunk_size can be inferred from result
-  type. In this example, chunk_size is 8.
+  lane level only accepts the offsets variant. Each lane loads a contiguous 1D
+  vector whose length is given by the result type; here each lane loads 8
----------------
Jianhui-Li wrote:

please update the example below to ensure that all shape must be shape: mask, offsets, and result. 

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