================
@@ -942,18 +906,35 @@ class CIRBaseBuilderTy : public mlir::OpBuilder {
return cir::CmpOp::create(*this, loc, kind, lhs, rhs, fenv);
}
+ /// Compares \p lhs and \p rhs element-wise. Each result element is a
+ /// signed integer of the operand element's bit width, 80 for x87 long
+ /// double.
cir::VecCmpOp createVecCompare(mlir::Location loc, cir::CmpOpKind kind,
----------------
andykaylor wrote:
Do we really still need this? There are just a few remaining users in builtins,
and I'd be surprised if we could get the expected type from the expression
there. The fact that we're saying we will produce a vector of 80-bit integers
for a vector of FP80 (which is technically different than "x87 long double",
BTW) is kind of weird. I really don't think we'll ever want that.
https://github.com/llvm/llvm-project/pull/230302
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
