[Mlir-commits] [mlir] [mlir][x86vector] AVX512-BF16 Dot op (PR #124800)

[Renato Golin]

================
@@ -271,6 +271,94 @@ def Vp2IntersectQIntrOp : AVX512_IntrOp<"vp2intersect.q.512", 2, [
                    VectorOfLengthAndType<[8], [I64]>:$b);
 }
 
+//===----------------------------------------------------------------------===//
+// AVX512-BF16 op definitions
+//===----------------------------------------------------------------------===//
+
+// Operation that is part of the input dialect.
+class AVX512BF16_Op<string mnemonic, list<Trait> traits = []> :
+  Op<X86Vector_Dialect, "avx512bf16." # mnemonic, traits> {}
----------------
rengolin wrote:

This is my point. Intel went a bit crazy with `avx512*` variations, and perhaps we should just collapse them all into the same. I'd start that way, and if we find clash problems, we can create a new sub division (`avx512.bf16`) or something.

https://github.com/llvm/llvm-project/pull/124800