https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121587
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> --- This might be better(?) done during vector lowering on GIMPLE as the alternative is for the target to claim to support FP16 add/sub/mult/div. I also wonder whether one can efficiently emulate FP16 rounding/truncation on floats or whether the actual conversion roundtrip is more efficient.
