[Mlir-commits] [mlir] [MLIR][XeGPU] Fix blocking pass for scf.if distribution (PR #207060)

[Jianhui Li]

================
@@ -924,6 +924,21 @@ xegpu::precomputeLoopBlockArgTypes(Operation *topLevelOp,
         recordTypes(init, {arg, res, yieldVal});
       return;
     }
+    if (auto ifOp = dyn_cast<scf::IfOp>(op)) {
+      // scf.if results expand 1:N during blocking. Record each result with the
----------------
Jianhui-Li wrote:

please update the comments before the function to include IfOp. I would avoid too detailed comment if they are really necessary but try to maintain the high level summary comment to be consistent. 

https://github.com/llvm/llvm-project/pull/207060