[Bug tree-optimization/126465] [16/17 Regression] Wrong code with IFN_COND_FMA folding on SVE

[tnfchris at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126465

--- Comment #3 from Tamar Christina <tnfchris at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #2)
> The else values, that is.

Yes, I am testing a patch.

> I wonder whether the folding is then still
> useful?  Maybe for constant else values, aka (negate! @5).

negate is not part of the forwarding region on FMA chains, so negating the
argument rather than the result gets the FMA to still forward the result to the
next one.

For for CPUs without this forwarding anyway it's not any worse.