https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124180

--- Comment #14 from rguenther at suse dot de <rguenther at suse dot de> ---
On Mon, 23 Feb 2026, tnfchris at gcc dot gnu.org wrote:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124180
> 
> --- Comment #12 from Tamar Christina <tnfchris at gcc dot gnu.org> ---
> (In reply to Richard Biener from comment #11)
> > (In reply to Richard Biener from comment #8)
> > ...
> > RVV supports same size, double size and half size conversions.  I expect
> > SVE to do the same, so arbitrary conversion optabs might be unneeded but
> > then if-conversion might need to put inbetween conversions in and
> > possibly pre-select between cond_{unpack,pack,}_fix{uns,} given it performs
> > a (premature) vectorization check on the chosen IFN.
> 
> We do support those as well indeed.  I was thinking that one option would be
> for ifcvt to generate the cond_ IFN regardless of target support and then we
> have vectorizable_conversion use this as more of a hint that we want a
> conversion but it still has to figure out if the target supports masking and 
> to
> which size.
> 
> This would perhaps allow us not to have to duplicate the logic into ifcvt and
> still only deal with it in vectorizable_conversion?

Yes, I agree we should avoid duplicating this.  We probably want to
remove the target check from if-conversion in general and only keep
some simple overall check in the future.

Reply via email to