[Mlir-commits] [mlir] [MLIR][XeGPU] Improve multi-reduction layout inference (PR #224618)

[Jianhui Li]

================
@@ -2493,13 +2458,39 @@ xegpu::SliceAttr xegpu::setupMultiReductionResultLayout(
                  "source shape not divisible by sg_layout");
           sgData[i] = srcShape[i] / sgLayout[i];
           remainingSgCount /= sgLayout[i];
-          order[i] = remainOrder++;
         }
       }
-      DenseI32ArrayAttr resOrderAttr = DenseI32ArrayAttr::get(
-          context, SmallVector<int32_t>(order.begin(), order.end()));
-      if (!orderAttr || orderAttr.empty())
-        resOrderAttr = nullptr;
+
+      DenseI32ArrayAttr resOrderAttr = nullptr;
----------------
Jianhui-Li wrote:

nit: please add a comment here to describe the logic below is to set orderAttr so that reductionDims are in row-major contiguous with its left-hand neighbor

https://github.com/llvm/llvm-project/pull/224618