[Mlir-commits] [mlir] [mlir][x86] Add pattern to tile and unroll vector.contract (PR #229076)

[Adam Siemieniuk]

================
@@ -0,0 +1,34 @@
+//===-- X86Enums.td - X86 dialect enum definitions ---------*- tablegen -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file defines the enums of the X86 dialect.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef X86_ENUMS
+#define X86_ENUMS
+
+include "mlir/IR/EnumAttr.td"
+
+//===----------------------------------------------------------------------===//
+// MLUTarget
+//===----------------------------------------------------------------------===//
+
+// The target CPU feature for the vector.contract multi-level unroll pattern
+// determines native shapes and supported data types for operands and the
+// accumulator, and corresponds to one of the nanokernel patterns.
+def MLUTarget : I32EnumAttr<"MLUTarget",
+    "Target CPU feature for the vector.contract multi-level unroll pattern", [
+  I32EnumAttrCase<"AMX_BF16", 0, "amx-bf16">, //-> VectorContractToAMXDotProduct
+  I32EnumAttrCase<"AMX_INT8", 1, "amx-int8">, //-> VectorContractToAMXDotProduct
----------------
adam-smnk wrote:

+1 for the enum but if we're already going this route, I'd generalize it. Sth toward optimization target or sth.

https://github.com/llvm/llvm-project/pull/229076