[Mlir-commits] [mlir] [MLIR][XeGPU] Add Layout Propagation support for multi-reduction/reduction op with scalar result (PR #189133)

[Jianhui Li]

================
@@ -1321,6 +1391,21 @@ LogicalResult ResolveLayoutConflicts::run() {
   return r.wasInterrupted() ? failure() : success();
 }
 
+LogicalResult
+ResolveLayoutConflicts::assignScalarResultLayout(OpResult &result) {
+  Operation *ProducerOp = result.getDefiningOp();
+  // Get the current layout of the vector value.
+  auto producerLayout = xegpu::getDistributeLayoutAttr(result);
+  // Insert a convert_layout op to resolve the conflict.
+  builder.setInsertionPointAfterValue(result);
+  auto convertOp = xegpu::ConvertLayoutOp::create(
----------------
Jianhui-Li wrote:

added

https://github.com/llvm/llvm-project/pull/189133