[Bug middle-end/124491] [14 Regression] ICE in fold_vec_perm during ccp with -msve-vector-bits=256

[rguenther at suse dot de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124491

--- Comment #13 from rguenther at suse dot de <rguenther at suse dot de> ---
On Wed, 25 Mar 2026, acoplan at gcc dot gnu.org wrote:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124491
> 
> --- Comment #12 from Alex Coplan <acoplan at gcc dot gnu.org> ---
> I tested some backports to the 14 branch and can confirm they fix the issue
> there (and survive bootstrap/regtest on aarch64).  For GCC 14, however, we also
> need Tamar's r15-4462-g55f898008ec8235897cf56c89f5599c3ec1bc963 to be
> backported as well, since the changes to tree-vect-generic.cc:lower_vec_perm in
> r16-6709 depend on that commit.
> 
> I checked that the ICE fixed by r15-4462 also shows up on the 14 branch with
> the same testcase, so it does seem worth backporting the dependent patch.
> 
> OK to backport the following revs to the 14 branch:
> 
> r15-4462-g55f898008ec8235897cf56c89f5599c3ec1bc963
> r16-6709-ga4716ece529dfd29d169ccc96979f7c747231f25
> r16-6743-ge787d5ace5cc315892868f3e5e2a2a16fd31bcc8
> r16-8111-g11f073a805d695d2dd4429530429ebf1561b4810
> 
> ?

OK.