[Mlir-commits] [mlir] [MLIR][XeGPU]Extend load_matrix/store_matrix to support 1D SLM access (PR #198652)

[Jianhui Li]

================
@@ -1308,23 +1319,28 @@ def XeGPU_StoreMatrixOp: XeGPU_Op<"store_matrix", [MemoryEffects<[MemWrite]>,
   let assemblyFormat = [{ $data `,` $mem_desc `` custom<DynamicIndexList>($offsets, $const_offsets)
                           prop-dict attr-dict `` `:` type(operands)}];
   let description = [{
-    This operation stores an nD `data` fragment into the shared local memory region
-    specified by an nD `mem_desc`. Memory descriptors of any rank are supported.
+    This operation stores a `data` fragment into the shared local memory region
+    specified by a `mem_desc`. Memory descriptors of any rank (1D or higher) are supported.
 
     This operation serves as an anchor through which users assign a layout attribute
     to govern computation distribution.
 
     Arguments:
      - `mem_desc`: the memory descriptor specifying the SLM region.
-     - `offsets`: the coordinates within the matrix where the data will be written.
-     - `data`: the values to be stored in the matrix.
+     - `offsets`: the coordinates within the memory descriptor where the data will be written.
+     - `data`: the values to be stored.
      - `subgroup_block_io`: [optional] An attribute indicating that the operation can be lowered
----------------
Jianhui-Li wrote:

The subgroup_block_io is still valid with 1D case. 

https://github.com/llvm/llvm-project/pull/198652