Re: [PATCH v2] s390x: Optimize vector permute with constant indexes

2024-04-09 Thread Juergen Christ
Am Tue, Apr 09, 2024 at 05:01:18PM +0200 schrieb Andreas Krebbel: > On 4/9/24 16:31, Juergen Christ wrote: > > Loop vectorizer can generate vector permutes with constant indexes > > where all indexes are equal. Optimize this case to use vector > > replicate instead of vector permute. > > > > gcc/

Re: [PATCH v2] s390x: Optimize vector permute with constant indexes

2024-04-09 Thread Andreas Krebbel
On 4/9/24 16:31, Juergen Christ wrote: > Loop vectorizer can generate vector permutes with constant indexes > where all indexes are equal. Optimize this case to use vector > replicate instead of vector permute. > > gcc/ChangeLog: > > * config/s390/s390.cc (expand_perm_as_replicate): Implem

[PATCH v2] s390x: Optimize vector permute with constant indexes

2024-04-09 Thread Juergen Christ
Loop vectorizer can generate vector permutes with constant indexes where all indexes are equal. Optimize this case to use vector replicate instead of vector permute. gcc/ChangeLog: * config/s390/s390.cc (expand_perm_as_replicate): Implement. (vectorize_vec_perm_const_1): Call new