[Mlir-commits] [mlir] [MLIR][XeGPU] Add sg_map attribute to support Work Item level semantics (PR #108864)

[Chao Chen]

================
@@ -77,12 +77,16 @@ def XeGPU_TensorDesc: XeGPUTypeDef<"TensorDesc", "tensor_desc",
 
     // A TensorDesc with 8x16 f32 elements for a memory region in shared memory space.
     xegpu.tensor_desc<8x16xf32, #xegpu.tdesc_attr<memory_scope = slm>>
+
+    // A TensorDesc with a sg_map
+    xegpu.tensor_desc<8x16xf32, #xegpu.sg_map<wi_layout = [1, 16], wi_data = [1, 1]>>
     ```
   }];
 
   let parameters = (ins ArrayRefParameter<"int64_t">: $shape,
                         "mlir::Type": $elementType,
-                        OptionalParameter<"mlir::Attribute">: $encoding);
+                        OptionalParameter<"mlir::Attribute">: $encoding,
+                        OptionalParameter<"mlir::Attribute">: $sg_map);
----------------
chencha3 wrote:

Is it also available to scattered TensorDesc (created by create_tdesc), which has slightly different semantics to Blocked TensorDesc, the one created by create_nd_tdesc? If so, is there any restrictions on it? 

https://github.com/llvm/llvm-project/pull/108864