[Mlir-commits] [mlir] [MLIR][XeGPU] Add offsets to load/store/prefetch (PR #150576)

Adam Siemieniuk llvmlistbot at llvm.org
Mon Jul 28 03:08:53 PDT 2025


================
@@ -189,6 +189,7 @@ def XeGPU_TensorDesc: XeGPUTypeDef<"TensorDesc", "tensor_desc",
   let genVerifyDecl = 1;
 }
 
+def XeGPU_TensorDescOrMemRef : AnyTypeOf<[XeGPU_TensorDesc,Non0RankedMemRefOf<[XeGPU_ScalarType]>, UI64]>;
----------------
adam-smnk wrote:

First, unrelated change to the PR. Could it be done separately so the offset addition is not blocked?

Second, why is it now allowed? What are the semantics? Why `load_nd` is now different?
It's basically undefined as is.

https://github.com/llvm/llvm-project/pull/150576


More information about the Mlir-commits mailing list