================
@@ -2521,6 +2521,32 @@ def ComplexImagOp : CIR_Op<"complex.imag", [Pure]> {
   let hasFolder = 1;
 }
 
+//===----------------------------------------------------------------------===//
+// ComplexAddOp
+//===----------------------------------------------------------------------===//
+
+def ComplexAddOp : CIR_Op<"complex.add", [Pure, SameOperandsAndResultType]> {
----------------
bcardosolopes wrote:

> I'm not sure why the decision was made to combine them for cir.binop. ... 
> bcardosolopes What is your opinion?

I kept them combined for simplicity given all uses so far don't really require 
them to be separate (and lowering is still as simple as if they were separate). 
In hindsight using tablegen class for binop would have had similar effect.

Whenever it makes sense to split off cir.binop (commutativity spec, etc as 
mentioned by @xlauko, or just plain consistency), I'm in favor. In that spirit, 
I'd say let's keep `complex` separate to mirror the dialect.

> Sidenote: I would not incline to complex dialect that much, as it might get 
> deprecated, since there is no maintainer still. See: 
> https://discourse.llvm.org/t/mlir-project-maintainers/87189

I had the same impression when I read this post. @andykaylor do any of your 
colleagues (or you) could be interested in maintaining the complex dialect?

https://github.com/llvm/llvm-project/pull/147578
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to