On Tue, 6 Jan 2026, Rainer Orth wrote:
> The gcc.dg/vect/slp-19c.c test FAILs on SPARC (32 and 64-bit):
>
> FAIL: gcc.dg/vect/slp-19c.c scan-tree-dump-times vect "vectorized 2 loops" 1
> FAIL: gcc.dg/vect/slp-19c.c scan-tree-dump-times vect "vectorizing stmts
> using SLP" 4
>
> The dump shows
>
> gcc.dg/vect/slp-19c.c:31:17: missed: unsupported vect permute { 1 3 }
> gcc.dg/vect/slp-19c.c:31:17: missed: permutation not supported, using
> elementwise access
> gcc.dg/vect/slp-19c.c:31:17: missed: not falling back to elementwise
> accesses
> gcc.dg/vect/slp-19c.c:40:25: missed: not vectorized: relevant stmt not
> supported: _16 = in[_15];
> gcc.dg/vect/slp-19c.c:31:17: note: unsupported SLP instance starting from:
> ia[i_136] = _16;
>
> so the test needs to require vect_perm.
>
> Tested on sparc-sun-solaris2.11 and i386-pc-solaris2.11.
>
> Ok for trunk?
I think it's only vect_extract_even_odd that's required.
OK to add that.
Richard.
>
> Rainer
>
>
--
Richard Biener <[email protected]>
SUSE Software Solutions Germany GmbH,
Frankenstrasse 146, 90461 Nuernberg, Germany;
GF: Jochen Jaser, Andrew McDonald, Werner Knoblich; (HRB 36809, AG Nuernberg)
