================
@@ -4002,8 +4012,11 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl
GD, unsigned BuiltinID,
Result = Builder.CreateBinaryIntrinsic(
Ty->isSignedIntegerType() ? Intrinsic::smin : Intrinsic::umin, Op0,
Op1, nullptr, "elt.min");
- } else
- Result = Builder.CreateMinNum(Op0, Op1, /*FMFSource=*/nullptr,
"elt.min");
+ } else {
+ FastMathFlags FMF;
+ FMF.setNoSignedZeros(true);
----------------
wzssyqa wrote:
> But they are in fact, different operations. And you have many choices for
> which FP min/max.
>
> Given this name doesn't have the historic fmin/fmax in it, I don't think this
> should take the fuzzy signed zero handling
You are right, since we have
`__builtin_elementwise_max/__builtin_elementwise_min`, we should not add `nsz`
to
`maxnum/minnum`.
https://github.com/llvm/llvm-project/pull/113133
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
