================
@@ -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:
> I agree this should not continue to use minnum/maxnum, though that change
> should not be part of this PR.
>
So can we merge this PR?
> minnum/maxnum behavior has never been consistent across targets (or across
> scenarios).
In fact, all of the architectures that claims implement IEEE754-2008, has the
same behavior:
AArch64, MIPSr6, LoongArch, PowerPC/VSX
That's why I'd plan to define `minnum/maxnum` as the same as these
architectures.
> But we could also add
> __builtin_elementwise_maximumnum/__builtin_elementwise_minimnumnum
I will do it.
https://github.com/llvm/llvm-project/pull/113133
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
