On Wed, 7 Jan 2026, Rainer Orth wrote:

> Hi Richard,
> 
> > On Tue, 6 Jan 2026, Rainer Orth wrote:
> >
> >> Two tests FAIL on 32 and 64-bit SPARC:
> >> 
> >> FAIL: gcc.dg/vect/vect-bool-2.c scan-tree-dump vect "optimized: loop 
> >> vectorized"
> >> FAIL: gcc.dg/vect/vect-bool-cmp-4.c scan-tree-dump vect "optimized: loop 
> >> vectorized"
> >> 
> >> The dump shows
> >> 
> >> gcc.dg/vect/vect-bool-2.c:12:12: missed:   not vectorized: relevant stmt 
> >> not supported: y_7 = _1 <= 16;
> >> 
> >> so the test needs vect_bool_cmp.
> >
> > <= 16 suggests a comparison in type char, not bool tough.  There's no
> > effective target for compares in general it seems?  But indeed we
> > might read that as "QImode compare"?  Given that's what
> > vect-bool-cmp-2.c requires for example.  That also requires
> > bool-cmp mask to bool vec_cond_mask.
> >
> > It would be so nice if the effective target names would map to
> > optabs (or combinations of those).
> 
> very true: unless you're intimately familar with vectorization, it's
> quite some guesswork which effective-target to use currently.

And even the vectorizer dump could be improved to mention relevant
missed optabs (sometimes it's multiple ones or one of many, of course...)

Richard.

Reply via email to