[Mlir-commits] [mlir] [MLIR][XeGPU] Introduce `xegpu::uArch` usage in target-sensitive passes (PR #163801)

[Artem Kroviakov]

================
@@ -456,7 +482,22 @@ void LayoutInfoPropagation::visitPrefetchNdOp(
   // Here we assign the default layout to the tensor descriptor operand of
   // prefetch.
   auto tdescTy = prefetch.getTensorDescType();
-  auto prefetchLayout = getDefaultSIMTLayoutInfo(tdescTy);
+
+  auto uArch = getUArch(getChipStr(prefetch).value_or(""));
----------------
akroviakov wrote:

Added `llvm_unreachable` to `getUArch()`. If we call the function, then we definitely care about the returned object in any place of the codebase, and it must be a valid/known uarch.

https://github.com/llvm/llvm-project/pull/163801