[Mlir-commits] [mlir] [mlir][ArmSME] Support lowering masked vector.outerproduct ops to SME (PR #69604)
Benjamin Maxwell
llvmlistbot at llvm.org
Wed Oct 25 07:26:36 PDT 2023
================
@@ -79,6 +79,23 @@ def ArmSME_TileSliceLayoutAttr : EnumAttr<ArmSME_Dialect, TileSliceLayout,
let defaultValue = "TileSliceLayout::Horizontal";
}
+def CombiningKind : I32EnumAttr<"CombiningKind", "Kind of combining function", [
+ I32EnumAttrCase<"Add", 0, "add">,
+ I32EnumAttrCase<"Sub", 1, "sub">,
+]> {
+ let cppNamespace = "::mlir::arm_sme";
+ let genSpecializedAttr = 0;
+}
+
+/// An attribute that specifies how to combine a newly produced value with the
+/// accumulator. This is similar to vector::CombiningKindAttr, but limited to
----------------
MacDue wrote:
I did try this, but including `VectorAttributes.td` within this file results in the generator creating duplicate definitions of attributes from the vector dialect (which fails to build).
https://github.com/llvm/llvm-project/pull/69604
More information about the Mlir-commits
mailing list